Particle Acceleration and Detection

François Méot

Understanding
the Physics
of Particle
Accelerators
A Guide to Beam Dynamics Simulations
Using ZGOUBI
Particle Acceleration and Detection

Series Editors
Alexander Chao, SLAC, Stanford University, Menlo Park, CA, USA
Katsunobu Oide, KEK, High Energy Accelerator Research Organization, Tsukuba,
Japan
Werner Riegler, Detector Group, CERN, Geneva, Switzerland
Frank Zimmermann, BE Department, ABP Group, CERN, Genèva, Switzerland

Editorial Board
Vladimir Shiltsev, Accelerator Physics Center, Fermilab, Batavia, IL, USA
The series “Particle Acceleration and Detection” is devoted to monograph texts
dealing with all aspects of particle acceleration and detection research and advanced
teaching. The scope also includes topics such as beam physics and instrumentation as
well as applications. Presentations should strongly emphasize the underlying physical
and engineering sciences. Of particular interest are
– contributions which relate fundamental research to new applications beyond the
immediate realm of the original field of research
– contributions which connect fundamental research in the aforementioned fields
to fundamental research in related physical or engineering sciences
– concise accounts of newly emerging important topics that are embedded in a
broader framework in order to provide quick but readable access of very new
material to a larger audience
The books forming this collection will be of importance to graduate students and
active researchers alike.
François Méot

Understanding the Physics

of Particle Accelerators
A Guide to Beam Dynamics Simulations
Using ZGOUBI
François Méot
Department of Collider-Accelerator
Brookhaven National Laboratory
Upton, NY, USA

ISSN 1611-1052 ISSN 2365-0877 (electronic)

Particle Acceleration and Detection
ISBN 978-3-031-59978-1 ISBN 978-3-031-59979-8 (eBook)
https://doi.org/10.1007/978-3-031-59979-8

© The Editor(s) (if applicable) and The Author(s) 2024. This book is an open access publication.

Open Access This book is licensed under the terms of the Creative Commons Attribution 4.0 International
License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribu-
tion and reproduction in any medium or format, as long as you give appropriate credit to the original
author(s) and the source, provide a link to the Creative Commons license and indicate if changes were
made.
The images or other third party material in this book are included in the book’s Creative Commons license,
unless indicated otherwise in a credit line to the material. If material is not included in the book’s Creative
Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted
use, you will need to obtain permission directly from the copyright holder.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, expressed or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland

Paper in this product is recyclable.

To my wife Sylvaine for her patient support
Foreword

The author of this book, Dr. François Méot, is an accelerator physicist at Brookhaven
National Laborartory (BNL). Currently, he is promoting the acceleration of proton
beams that maintain spin polarization at BNL’s large accelerator RHIC, and the
FFAG accelerator project, especially the non-scaling FFAG type Energy Recovery
Linac (ERL) accelerator project. He has been developing for some time a numerical
simulation code ZGOUBI for analyzing the dynamical behavior of charged particles
including the spin motions in various types of accelerators.
Accelerators are devices that focus and accelerate charged particles using elec-
tromagnetic force, and have been developed as experimental tools for particle and
nuclear physics as well as for various applications. Various types of accelerators have
been invented as described in each chapter of this book. In order to understand the
behaviors of the charged particle motion in these accelerators, it is necessary to solve
the equation of motion using numerical methods and many numerical simulation
codes have been developed. In particular, in ring accelerators where closed orbits are
unknown in advance, such as scaling FFAG, it is necessary to accurately determine
particle trajectories using electromagnetic field distributions obtained through actual
measurements or numerical calculations. Also, when evaluating spin motion in a ring
accelerator, identification and distortion of closed orbits are important points. For
these reasons, there was a need for a simulation code that could evaluate the particle
orbits from beam tracking based on the external electromagnetic field distribution,
regardless of the type of accelerator. The code ZGOUBI answers those needs.
In the 1980s, pioneering work on spin-polarized proton beam acceleration in
proton synchrotrons began with the 3 GeV proton synchrotron SATURNE 2 in Saclay,
François. Dr. Méot contributed to polarized proton beam acceleration studies in
SATURNE 2 by extending the ZGOUBI code, which was originally developed to
analyze the orbits within magnetic field spectrometers, to the analysis of proton spin
motion in synchrotrons. In the recent years, he made a major contribution to the
success of polarized proton beam acceleration in RHIC.
In the late 1990s, I developed the world’s first proton FFAG accelerator (POP-
FFAG) with our colleagues at KEK, Japan. I first met Dr. Méot at an international

vii
viii Foreword

conference in Paris. We decided to start the collaboration on the FFAG experi-

mental study in Japan. Japan’s monsoon summer is extremely hot and humid, and I
remember the days spent doing beam experiments day after day. For François Méot,
the experience brought back memories of his childhood in Southeast Asia.

Ashiya, Japan Yoshiharu Mori

January 2024
Preface

“The history of SATURNE 1 and SATURNE 2 accelerators includes the construction

of 4 spectrometers as particle detectors. The first, SPES I designed in 1968 is of
high resolution. The resolution permits the selection of one nuclear level which is
necessary even for elastic scattering. A program, ZGOUBI (Jean Claude Faivre,
Denis Garreta) for precise orbit computing had to be written and was the first one
to be fast enough to permit new spectrometer design.”1
In the matter of raytracing accuracy, no method competes with stepwise numerical
integration of the differential equations of motion. This holds for beam dynamics
in accelerators and beamlines (which abide by the Lorentz force equation), for
spin precession (Thomas-BMT differential equation), computation of the radiation
Poynting vector, or when adding such ingredients as scattering and other stochastic
energy loss.
In the era of artificial intelligence (AI) and machine learning (ML), and leaning
on high performance computing (HPC) for speed and statistics, there is no reason
to deprive oneself of building informed intelligence on deep understanding of the
kinematics, by pushing charged particles step by step in true models of magnetic and
electric fields.
Computer codes for stepwise raytracing have long been available and used. Their
long lists of achievements include
– As early as the 1950s, the design of fixed field alternating gradient (FFAG) cyclic
accelerators by the MURA group, from paraxial parameters to dynamical accep-
tance performance, using a Runge-Kutta integrator concurrently with magnetic
field models including computed field maps.2
– The design and parameterization of nuclear physics mass spectrometers from the
early times, decades ago. This is the only way that numerical simulation can

1 Jacques Thirion, Preface, in “The 20 years of the synchrotron SATURNE 2”. A. Boudard, P.-A.
Chamouard Eds. World Scientific, May 1998.
2 Cole, F. T.: O Camelot, a Memoir of the MURA Years (April 11, 1994); Sects. 7.1, 10.5 and 10.6.

Available in the Proceedings of the Cyclotron Conference, East Lansing, USA, May 13–17, 2001.
https://accelconf.web.cern.ch/c01/cyc2001/extra/Cole.pdf.

ix
x Preface

match the level of time-of-flight and momentum resolution requested from these
devices.
– The simulation of radiation and bunch-radiation interaction, as part of the design
of insertion devices found in synchrotron radiation rings and free electron lasers.
– The analysis of spin diffusion and polarization lifetime in electron accelerators
and storage rings.
– Computation and correction of optical aberrations. They matter in all beam
optics problems: imaging and image resolution, geometrical acceptance in beam
lines and dynamical acceptance in rings, non-linear resonant extraction for
hadrontherapy purpose, etc.

For accuracy, the representation of electrostatic and/or magnetic fields may resort
to analytical modeling or to computed field maps, whereas full benefit from these
models requires stepwise integration of the equations of motion. An additional
benefit of the method is its allowing high resolution Monte Carlo processes, such as
synchrotron radiation, in-flight decay, particle-matter interaction, etc.
Various numerical integrators have been in use for raytracing, including Taylor
series, Runge-Kutta, leap-frog, etc. They are found in decades old, popular codes as
RAYTRACE, GEANT, ZGOUBI, and others. Accuracy on the modeling of fields,
allied with accuracy on the integration of the equations of motion, makes the numer-
ical quality of this partnership hard to beat—and these proven ‘old’ codes best modern
tools!
A drawback of the latter might be considered to be the longer time it takes to push
particles, compared to matrix and other kick-drift mapping techniques. Computing
speed is often used as a justification to jeopardize intelligence by resorting to approx-
imations regarding kinematics, or fields, or both. However (i) better do things slowly
and right than fast and wrong; (ii) with today’s HPC, only a few accelerator design
and operation simulation problems, e.g., space charge, dynamical aperture compu-
tation in large rings, remain affected by somewhat “long” execution times. Where
paraxial optics or quasi-linear optical systems are concerned, low order approxima-
tions might, why not, fulfill requirements and accuracy might be relaxed, for the sake
of computing speed ... however, several categories of beam dynamics simulations
in these paraxial machines (such as multiturn bunch tracking in small rings) may
anyway be performed in a reasonably short time, the more so using HPC and/or
CPU clusters, without having to surrender to mapping style of field modeling and
kinematics approximations.
In the era of AI and ML, there is no reason to rely on approximate integration
methods, and on approximate field models: loose modeling is in patent contradiction
with the goals of intelligence and learning! It would be akin to assuming π = 3.1416,
or c = 3 × 108 m/s in designing accelerators—this does not happen. Stepwise inte-
gration in realistic field models saves on the time spent figuring out and overcoming
the adverse effects of approximations: “Is what I observe due to my approximations,
or is it real?”. This allows time to be efficiently used to focus one’s energy where it
is needed, and of interest, i.e., the physics of phenomena and their understanding.
Preface xi

Table 1 Comparative advantages (+) and disadvantages (−) of numerical simulations and machine
operation

Numerical simulation Machine operation

Studying the physics of No limitation + + that’s where the truth is
phenomena
Explore exotic dynamics No limitation + − technological and principle
limitations
Details of particle dynamics No limitation + − requires very small beam
emittance
Speed Improves with time − + can’t be beaten!
Cost Essentially zero + − a lot

The truth lies in machine operation anyway. This is an additional good reason
for numerical methods to stay away from approximations, as they may offset appre-
hension of phenomena and hamper comparisons with operation outcomes. Learning
from machine operation may have limitations, compared with raytracing techniques,
learning from the latter has its limitations as well, both are complementary anyway
(Table 1).
This book is an introduction to the physics of particle accelerators based on beam
dynamics simulations, covering accelerator concepts developed over the past century.
It is as much about learning on particle accelerators, as it is about learning beam
dynamics simulation techniques and tools, real life accelerator design methods, and
simulation data production and treatment.
Simulation exercises are proposed, this is the “hands-on” side of the learning
method. Their material is based on real life design studies, which have often been
subject to tutorials in workshops and university teachings, or used in conference
publications and peer-review journal articles. A lot more than proposed in this book,
covering half a century of accelerator, spectrometer, and beam line design studies, can
be found in the sourceforge branch https://sourceforge.net/p/zgoubi/code/HEAD/
tree/branches/exemples/.
These diverse studies have most of the time been subject to laboratory tech. notes
and other publications, which for some can be found in the sourceforge branch https://
sourceforge.net/p/zgoubi/code/HEAD/tree/branches/publications/.
More zgoubi simulation material, guidance regarding the use of the code, its
keywords, and regarding its capabilities, can be found in a general manner in
– US DOE OSTI repository 3 https://www.osti.gov/.
– PR-AB and NIM A publications,

3At the time of these writtings, “https://www.osti.gov/search/semantic:zgoubi” produces “164

Search Results”.
xii Preface

– JACoW accelerator conference proceedings site https://www.jacow.org/Main/Pro

ceedings and its search tool.4
Performing numerical simulations is not quite as real as being at the command/
controls in an accelerator facility, yet it may sometimes be quite close. Computer
simulations closely reproduce beam manipulations done in accelerator control rooms.
In fact, machine operation has much to do with beam dynamics simulation, since both
make extensive use of computer models to reproduce basic beam dynamics and accel-
erator optics, beam monitoring, and control. Simulations use similar post-processing
tools, and deliver similar types of data: particle coordinates, bunch parameters, phase
space portraits, motion spectra, machine parameters, etc.
Computer simulations proposed here manipulate, guide, and accelerate parti-
cles and bunches, in most styles of particle accelerators devised since the 1920s.
In performing these simulation exercises, the reader will acquire an understanding
of charged particle beam optics, extend their knowledge of accelerator physics and
technology, and assess the use of one or the other of the existing beam handling
technologies for beam delivery in such or such particle beam application. The exer-
cises take the reader in a virtual world of accelerator and beam simulations on a
computer, in the way that accelerator physicists design these machines in their labo-
ratories, and play with them in control rooms. They allow the basic theoretical and
practical aspects of the main technological components of particle accelerators to
be discovered: guiding and focusing using E and/or B fields, accelerating in voltage
gaps, shaking particles in wiggler magnets, getting rid of impurities in combined E,
B devices, etc. Checking the consistency of numerical results against the elements
of theory introduced in the chapters, and vice-versa, is part of these exercises.
Finally, in some exercises one may think of launching a bunch tracking simulation
for minutes, and why not hours, while tuning the beam and taking data as the computer
is quietly pushing these bunches, repeatedly, through a virtual beam line, or around
a virtual ring. Along that line, bunch tracking animations are proposed in some
exercises, based on short beam lines or small rings for conveniently fast tracking.
Most accelerator species are covered in this book, in a series of chapters
ordered following the historical chronology: from electrostatic systems to today’s
storage rings, via the classical cyclotron, relativistic cyclotron, microtron, beta-
tron, synchrotron, FFAG. Without forgetting linacs, opportunistically, if not chrono-
logically, introduced in racetrack microtron and in linear FFAG recirculator exer-
cises. Note in passing, quite interestingly all these accelerator species, invented and
developed over a century, are still in use today, in one application or another.
All the chapters are organized in a similar manner:
– a short introduction with some historical insight,
– a brief theoretical reminder, which provides recipes resorted to in the exercises,
– a series of simulation exercises.

4At the time of these writtings, looking up ‘zgoubi’ in JACoW advanced search tool produces about
800 results.
Preface xiii

The solutions of the exercises are the subject of a dedicated chapter. They include
zgoubi input data files, or detailed indications to build them, as well as expected
results.
The reader is assumed to have a basic knowledge of charged particle beam optics
in transport lines and accelerators. So the theoretical reminders are rather concise,
aimed essentially at bringing the basic concepts and formulæ used in the exercises,
with minimal explanations. Thus, having text books at hand when working on the
simulations is a good idea. The reader may at times feel that computer code capa-
bilities are a little (too) lightly addressed ... well ... the 400 page companion to the
present opus, Zgoubi Users’ Guide, happens to be indispensable as well to work out
the simulations. It is available in its most recent version at https://sourceforge.net/p/
zgoubi/code/HEAD/tree/trunk/guide/Zgoubi.pdf.
All details regarding the methods to work it out may not be provided when
proposing a simulation exercise. For instance which keywords may be preferred (e.g.,
for a bend magnet: BEND, CARTEMES, DIPOLE, MULTIPOL, TOSCA, ...), which
output file(s) will deliver results in an appropriate form (e.g., zgoubi.res, zgoubi.plt,
zgoubi.fai, zgoubi.TWISS.out, zgoubi.MATRIX.out, ...), post-treatment to possibly
apply on these files content. Thinking about it and finding the preferred ways is part
of the exercise. Choices depend on the context, e.g., learning beam optics, teaching,
developing AI programs, or working in a team on a particular design. Of course,
this means a learning curve in order to figure out the possibilities offered by zgoubi
(and, beyond, by stepwise raytracing based beam optics techniques). This is one of the
goals here, and it is also why the exercise series starts with simple simulations using
a very limited number of keywords: in the first two cyclotron chapters for instance,
just DIPOLE or TOSCA, to bend a trajectory into a closed orbit in a magnetic field,
CAVITE for resonant acceleration, and FAISCEAU to monitor particle coordinates.
So... Yes! Grab your laptop, it will be an essential tool. You’ll be able to play with
charged particle beams in a world of virtual accelerator optical components, beam
lines, and rings, and ... have fun!

BNL, Upton, NY, USA François Méot

Acknowledgements

This book is a product of more than 40 years of numerical simulations within the
framework of tens of projects, in high energy physics and nuclear physics labora-
tories, CEA Saclay, BNL, CERN, FERMILAB, TRIUMF, and others. It is also the
product of years of tutorials, in workshops and university courses. Not the least, it
benefits from the feedbacks these collaborations, workshops, and teachings allowed.
More specifically this book has benefitted from discussions with my colleagues
and friends, among whom former PhD students, Yann Dutheil, Bhawin Dhital, Malek
Haj Tahar, Kiel Hock, Xiangdong Lee, Joseph Lidestri, Steve Peggs, Guillaume
Robert-Demolaize, Laurent Sérani, Victor Smirnov. They provided suggestions and
advice for corrections and improvements.
The work environment at the Collider-Accelerator Department at the Brookhaven
National Laboratory is for a large part responsible for this undertaking. I would like
to thank here the many colleagues with whom I have been in close contact during
these years of collaborations on accelerator R&D and projects at BNL C-AD, and in
particular Thomas Roser, Nick Tsoupas, Haixin Huang, Wolfram Fischer, and Dejan
Trbojevic.

xv
Contents

1 Numerical Simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Half a Century of Charged Particle Raytracing in Zgoubi . . . . . 2
1.2 Raytracing with Zgoubi—Solving the Exercises . . . . . . . . . . . . . 3
1.3 Graphics, Data Treatment: zpop, gnuplot, awk, python . . . . . . . . 9
1.4 Interface to Zgoubi? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2 Electrostatic Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.2 Basic Concepts and Formulæ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.2.1 Kinetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.2.2 Optical Components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.2.3 Periodic Structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.2.4 Spin Precession . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.3 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
Plane Condenser; Spin Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
Toroidal Condenser . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
A Time-of-Flight Mass Spectrometer Ring . . . . . . . . . . . . . . . . . . . 36
Converging Rays in a Bipotential Cylindrical Lens . . . . . . . . . . . . 38
The AGS Electron Analog . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
2.4 Solutions of Exercises of This Chapter: Electrostatic
Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
Plane Condenser; Spin Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
Toroidal Condenser . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
A Time-of-Flight Mass Spectrometer Ring . . . . . . . . . . . . . . . . . . . 42
Converging Rays in a Bipotential Cylindrical Lens . . . . . . . . . . . . 49
The AGS Electric Analog . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

xvii
xviii Contents

3 Classical Cyclotron . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
3.2 Basic Concepts and Formulæ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
3.2.1 Fixed-Energy Orbits, Revolution Period . . . . . . . . . . . . . . 62
3.2.2 Weak Focusing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
3.2.3 Quasi-Isochronous Resonant Acceleration . . . . . . . . . . . . 68
3.2.4 Beam Extraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
3.2.5 Spin Dance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
3.3 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
Modeling a Cyclotron Dipole: Using a Field Map . . . . . . . . . . . . . 73
Modeling a Cyclotron Dipole: Using an Analytical Field
Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
Resonant Acceleration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
Spin Dance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
Synchronized Spin Torque . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
Weak Focusing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
Loss of Isochronism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
Ion Trajectories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
RF Phase at the Accelerating Gap . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
The Cyclotron Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
Cyclotron Extraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
Acceleration and Extraction of a 6-D Polarized Bunch . . . . . . . . . 79
3.4 Solutions of Exercises of This Chapter . . . . . . . . . . . . . . . . . . . . . . 79
Modeling a Cyclotron Dipole: Using a Field Map . . . . . . . . . . . . . 79
Modeling a Cyclotron Dipole: Using an Analytical Field
Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
Resonant Acceleration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
Spin Dance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
Synchronized Spin Torque . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
Weak Focusing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
Loss of Isochronism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
Ion Trajectories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
RF Phase at the Accelerating Gap . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
The Cyclotron Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
Cyclotron Extraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
Acceleration and Extraction of a 6-D Polarized Bunch . . . . . . . . . 128
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
4 Relativistic Cyclotron . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
4.2 Basic Concepts and Formulæ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
4.2.1 Thomas Focusing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
4.2.2 Spiral Sector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
Contents xix

4.2.3 Isochronism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144

4.2.4 Cyclotron Extraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
4.2.5 Resonant Spin Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
4.3 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
Modeling Thomas AVF Cyclotron . . . . . . . . . . . . . . . . . . . . . . . . . . 147
Designing an Isochronous AVF Cyclotron . . . . . . . . . . . . . . . . . . . 148
Acceleration to 200 MeV in an AVF Cyclotron . . . . . . . . . . . . . . . 148
Thomas-BMT Spin Precession in Thomas Cyclotron . . . . . . . . . . 148
Isochronism and Edge Focusing in a Separated Sector
Cyclotron . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
A Model of PSI Ring Cyclotron Using CYCLOTRON . . . . . . . . . 150
4.4 Solutions of Exercises of This Chapter: Relativistic
Cyclotron . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
Modeling Thomas AVF Cyclotron . . . . . . . . . . . . . . . . . . . . . . . . . . 150
Designing an Isochronous AVF Cyclotron . . . . . . . . . . . . . . . . . . . 160
Acceleration to 200 MeV in an AVF Cyclotron . . . . . . . . . . . . . . . 161
Thomas-BMT Spin Precession in Thomas Cyclotron . . . . . . . . . . 166
Isochronism and Edge Focusing in a Separated Sector
Cyclotron . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
A Model of PSI Ring Cyclotron Using CYCLOTRON . . . . . . . . . 176
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
5 Betatron . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
5.2 Basic Concepts and Formulæ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
5.2.1 Betatron Condition. Acceleration . . . . . . . . . . . . . . . . . . . 191
5.2.2 Transverse Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
5.2.3 Synchroton Radiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
5.3 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
Develop a Betatron Magnet in Zgoubi . . . . . . . . . . . . . . . . . . . . . 195
A 315 MeV Betatron . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
Acceleration with Radiation Loss in the 315 MeV Betatron . . . . . 196
5.4 Solutions of Exercises of This Chapter: Betatron . . . . . . . . . . . . . . 196
Develop a Betatron Magnet in Zgoubi . . . . . . . . . . . . . . . . . . . . . 196
A 315 MeV Betatron . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
Acceleration with Radiation Loss in the 315 MeV Betatron . . . . . 202
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
6 Microtron . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
6.2 Basic Concepts and Formulæ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
6.2.1 Classical Microtron . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
6.2.2 Racetrack Microtron . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
6.2.3 Synchrotron Radiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
xx Contents

6.3 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215

Build a Classical Microtron . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
Build a 100 MeV Racetrack Microtron . . . . . . . . . . . . . . . . . . . . . . 216
6.4 Solutions of Exercises of This Chapter: Microtron . . . . . . . . . . . . 216
Build a Classical Microtron . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
Build a 100 MeV Racetrack Microtron . . . . . . . . . . . . . . . . . . . . . . 221
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
7 Synchrocyclotron . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
7.2 Basic Concepts and Formulæ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
7.2.1 Phase Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
7.2.2 Transverse Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
7.3 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
Operate a Cyclotron Dipole in Synchrocyclotron Mode . . . . . . . . 231
Operate an FFAG in Synchrocyclotron Mode . . . . . . . . . . . . . . . . . 231
7.4 Solutions of Exercises of This Chapter: Synchrocyclotron . . . . . . 231
Operate a Cyclotron Dipole in Synchrocyclotron Mode . . . . . . . . 231
Operate an FFAG in Synchrocyclotron Mode . . . . . . . . . . . . . . . . . 235
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
8 Weak Focusing Synchrotron . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
8.2 Basic Concepts and Formulæ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242
8.2.1 Periodic Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
8.2.2 Acceleration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
8.2.3 Synchrotron Radiation Poynting . . . . . . . . . . . . . . . . . . . . 258
8.2.4 Depolarizing Resonances . . . . . . . . . . . . . . . . . . . . . . . . . . 262
8.3 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 268
Construct SATURNE 1 Synchrotron. Spin Resonances . . . . . . . . 268
Construct the ZGS Synchrotron. Spin Resonances . . . . . . . . . . . . . 271
Visible SR from GEC 70 MeV Synchrotron . . . . . . . . . . . . . . . . . . 275
8.4 Solutions of Exercises of This Chapter: Weak Focusing
Synchrotron . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275
Construct SATURNE 1 Synchrotron. Spin Resonances . . . . . . . . 275
Construct the ZGS Synchrotron. Spin Resonances . . . . . . . . . . . . . 299
Visible SR from GEC 70 MeV Synchrotron . . . . . . . . . . . . . . . . . . 315
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 317
9 Strong Focusing Synchrotron . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 321
9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323
9.2 Basic Concepts and Formulæ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 328
9.2.1 Components of the Strong Focusing Optics . . . . . . . . . . . 329
9.2.2 Transverse Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 332
9.2.3 Resonances . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 337
9.2.4 Acceleration. Synchrotron Motion . . . . . . . . . . . . . . . . . . . 340
Contents xxi

9.2.5 Synchrotron Radiation, Dynamical Effects . . . . . . . . . . . 343

9.2.6 Visible Synchrotron Radiation. Interference . . . . . . . . . . 345
9.2.7 Polarization, Resonances . . . . . . . . . . . . . . . . . . . . . . . . . . . 347
9.3 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 349
Construct SATURNE 2 Synchrotron . . . . . . . . . . . . . . . . . . . . . . . . 349
Non-Linear Motion in SATURNE 2 . . . . . . . . . . . . . . . . . . . . . . . . . 351
SVD Orbit Correction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 352
Cornell Electron RCS. Radiative Energy Loss . . . . . . . . . . . . . . . . 352
Coupling in a Light Source Storage Ring . . . . . . . . . . . . . . . . . . . . 353
SR Electric Impulse and Interference in a Miniwiggler . . . . . . . . . 355
Depolarizing Resonances in SATURNE 2 . . . . . . . . . . . . . . . . . . . . 355
Ion and Electron Polarization. Preservation of Polarization . . . . . 357
9.4 Solutions of Exercises of This Chapter: Strong Focusing
Synchrotron . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 358
Construct SATURNE 2 Synchrotron . . . . . . . . . . . . . . . . . . . . . . . . 358
Non-Linear Motion in SATURNE 2 . . . . . . . . . . . . . . . . . . . . . . . . . 364
SVD Orbit Correction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 368
Cornell Electron RCS. Radiative Energy Loss . . . . . . . . . . . . . . . . 369
Coupling in a Light Source Storage Ring . . . . . . . . . . . . . . . . . . . . 373
SR Electric Impulse and Interference in a Miniwiggler . . . . . . . . . 374
Depolarizing Resonances in SATURNE 2 . . . . . . . . . . . . . . . . . . . . 376
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 380
10 FFAG, Scaling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 385
10.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 387
10.2 Basic Concepts and Formulæ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 389
10.2.1 Radial Sector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 393
10.2.2 Spiral Sector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 393
10.2.3 Longitudinal Motion. Acceleration . . . . . . . . . . . . . . . . . . 398
10.3 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 401
10.3.1 A 150 MeV, Proton, Radial Sector FFAG . . . . . . . . . . . . . 401
Field in a Radial Sector Dipole Triplet . . . . . . . . . . . . . . . . . . . . . . . 401
Orbits, Scalloping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 402
Zero-Chromaticity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 403
Beam Envelopes; Phase Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 403
Acceleration. Transverse Betatron Damping . . . . . . . . . . . . . . . . . . 404
10.3.2 RACCAM Proton Therapy Spiral Sector FFAG . . . . . . . 404
Field in a Spiral Sector Dipole . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 405
Orbits, Scalloping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 406
Zero-Chromaticity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 406
Beam Envelopes, Optical Functions . . . . . . . . . . . . . . . . . . . . . . . . . 407
Periodic Stability Domain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 407
Motion Stability Limit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 407
Dynamic Aperture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 407
Acceleration. Transverse Betatron Damping . . . . . . . . . . . . . . . . . . 407
xxii Contents

10.3.3 FFAG Acceleration Methods . . . . . . . . . . . . . . . . . . . . . . . 408

Hybrid Acceleration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 408
Bucket Acceleration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 408
Serpentine Acceleration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 409
10.4 Solutions of Exercises of This Chapter: FFAG, Scaling . . . . . . . . 409
10.4.1 A 150 MeV, Proton, Radial Sector FFAG . . . . . . . . . . . . . 409
Field in a Radial Sector Dipole Triplet . . . . . . . . . . . . . . . . . . . . . . . 409
Orbits, Scalloping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 412
Zero-Chromaticity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 415
Beam Envelopes; Phase Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 419
Acceleration. Transverse Betatron Damping . . . . . . . . . . . . . . . . . . 421
10.4.2 RACCAM Proton Therapy Spiral Sector FFAG . . . . . . . 426
Field in a Spiral Sector Dipole . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 426
Orbits, Scalloping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 429
Zero-Chromaticity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 433
Beam Envelopes, Optical Functions . . . . . . . . . . . . . . . . . . . . . . . . . 436
Periodic Stability Domain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 438
Motion Stability Limit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 438
Dynamic Aperture Scan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 439
Acceleration, Transverse Betatron Damping . . . . . . . . . . . . . . . . . . 440
10.4.3 FFAG Acceleration Methods . . . . . . . . . . . . . . . . . . . . . . . 441
Hybrid Acceleration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 441
Bucket Acceleration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 441
Serpentine Acceleration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 441
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 441
11 FFAG, Linear . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 445
11.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 446
11.2 Basic Concepts and Formulæ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 450
11.2.1 Linear FFAG Cell . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 450
11.2.2 Quasi-Isochronous Serpentine Acceleration . . . . . . . . . . . 452
11.2.3 Synchrotron Radiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 454
11.2.4 Polarization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 455
11.3 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 457
EMMA Ring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 457
eRHIC ERL FFAG2 loop. Synchrotron Radiation.
Polarization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 459
eRHIC 22 GeV ERL, From Start to End . . . . . . . . . . . . . . . . . . . . . 460
11.4 Solutions of Exercises of This Chapter: FFAG, Linear . . . . . . . . . 461
EMMA Ring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 461
eRHIC ERL FFAG2 loop. Synchrotron Radiation.
Polarization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 464
eRHIC 22 GeV ERL, From Start to End . . . . . . . . . . . . . . . . . . . . . 471
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 473
Contents xxiii

12 Beam Lines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 477

12.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 477
12.2 Basic Concepts and Formulæ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 478
12.2.1 Beam Expander . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 478
12.2.2 Nano-Probe Final Focus Achromat . . . . . . . . . . . . . . . . . . 482
12.2.3 A Muon Collect Channel . . . . . . . . . . . . . . . . . . . . . . . . . . 483
12.2.4 Low Energy Spin Rotator . . . . . . . . . . . . . . . . . . . . . . . . . . 492
12.2.5 Synchrotron Radiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 494
12.3 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 496
12.3.1 Beam Expander . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 496
Construct a Beam Expander Line . . . . . . . . . . . . . . . . . . . . . . . . . . . 497
Expander Line Optics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 499
Transverse Uniformization; Beam Scan . . . . . . . . . . . . . . . . . . . . . . 499
Animation: Transverse Scan by a Uniform Beam . . . . . . . . . . . . . 500
12.3.2 Nano-Probe E, B Final Doublet Achromat . . . . . . . . . . . . 501
Construct a Final Focus Doublet . . . . . . . . . . . . . . . . . . . . . . . . . . . . 501
E, B Lens, Hard-Edge Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 502
E, B Lens With Fringe Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 502
Animation: Dynamical Squeeze of Point-Spread Function . . . . . . 502
12.3.3 π → μ + ν In-Flight Decay, Bunch Densities . . . . . . . . 503
Parent Pion Bunch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 503
Muon Bunch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 503
12.3.4 Pion Funnel and Muon FODO Collect Channel . . . . . . . 504
Beam Line Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 505
Bunch Densities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 505
12.3.5 Low Energy Spin Rotator . . . . . . . . . . . . . . . . . . . . . . . . . . 506
Spin Rotation, Spin Matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 506
Integration Step Size . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 507
Add Fringe Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 507
Transport of a 6D Polarized Bunch . . . . . . . . . . . . . . . . . . . . . . . . . . 507
12.3.6 Synchrotron Radiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 508
Synchrotron Radiation in a Long FFAG Beam Line . . . . . . . . . . . 508
12.4 Solutions of Exercises of This Chapter: Beam Lines . . . . . . . . . . 508
12.4.1 Beam Expander . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 508
Construct a Beam Expander Line . . . . . . . . . . . . . . . . . . . . . . . . . . . 508
Expander Line Optics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 509
Transverse Uniformization; Beam Scan . . . . . . . . . . . . . . . . . . . . . . 515
Animation: Transverse Scan by a Uniform Beam . . . . . . . . . . . . . 519
12.4.2 Nano-Probe E, B Final Doublet Achromat . . . . . . . . . . . . 520
Construct a Final Focus Doublet . . . . . . . . . . . . . . . . . . . . . . . . . . . . 520
E, B Lens, Hard-Edge Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 525
E, B Lens, With Fringe Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 528
Animation: Dynamical Squeeze of Point-Spread Function . . . . . . 532
12.4.3 π →μ + ν In-Flight Decay, Bunch Densities . . . . . . . . . . 532
Parent Pion Bunch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 532
xxiv Contents

Muon Bunch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 534

12.4.4 Pion Funnel and Muon FODO Collect Channel . . . . . . . 542
Beam Line Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 542
Bunch Densities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 546
12.4.5 Low Energy Spin Rotator . . . . . . . . . . . . . . . . . . . . . . . . . . 551
Spin Rotation, Spin Matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 551
Integration Step Size . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 554
Add Fringe Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 555
Transport of a 6D Polarized Bunch . . . . . . . . . . . . . . . . . . . . . . . . . . 562
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 565
13 Spectrometer; Mass Separator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 567
13.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 568
13.2 Basic Concepts and Formulæ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 569
13.3 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 576
A Nuclear Physics Spectrometer, SPES II . . . . . . . . . . . . . . . . . . . . 577
SPES III Spectrometer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 578
A High-Resolution Mass Separator . . . . . . . . . . . . . . . . . . . . . . . . . 581
13.4 Solutions of Exercises of This Chapter: Spectrometer;
Mass Separator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 583
A Nuclear Physics Spectrometer, SPES II . . . . . . . . . . . . . . . . . . . . 583
SPES III Spectrometer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 592
A High-Resolution Mass Separator . . . . . . . . . . . . . . . . . . . . . . . . . 593
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 603
14 Optical Elements and Keywords, Complements . . . . . . . . . . . . . . . . . . 605
14.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 605
14.2 Drift Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 606
14.3 Guiding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 607
14.3.1 Dipole Magnet, Curved . . . . . . . . . . . . . . . . . . . . . . . . . . . . 608
14.3.2 Dipole Magnet, Straight . . . . . . . . . . . . . . . . . . . . . . . . . . . 609
14.3.3 Fringe Field, Modeling, Overlapping . . . . . . . . . . . . . . . . 610
14.3.4 Toroidal Condenser . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 613
14.4 Focusing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 613
14.4.1 Wedge Focusing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 613
14.4.2 Quadrupole . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 616
14.4.3 Solenoid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 618
14.5 Data Treatment Keywords . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 620
14.5.1 Concentration Ellipse: FAISCEAU, FIT[2],
MCOBJET, ... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 620
14.5.2 Transport Coefficients: MATRIX, OPTICS,
TWISS, Etc. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 621
Contents xxv

14.6 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 624

Magnetic Sector Dipole . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 624
Quadrupole Doublet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 625
Solenoid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 625
14.7 Solutions of Exercises of This Chapter: Optical Elements
and Keywords, Complements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 626
Magnetic Sector Dipole . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 626
Quadrupole Doublet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 631
Solenoid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 632
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 633

Index - Zgoubi Optical Elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 635

Chapter 1
Numerical Simulations

Several of the numerical simulation exercises proposed in the following chapters

require step-by-step raytracing through the optical elements concerned, for diverse
reasons, depending on the problem:
– some optical elements feature complicated magnetic or electrostatic fields,
– some are represented by magnetic and/or electrostatic field maps,
– fields may be time-dependent and vary while traversed,
– some processes require accuracy on field modeling: spin transport for instance, or
computation of the synchrotron radiation Poynting vector from the coordinates of
a relativistic particle,
– some Monte Carlo processes may require small integration steps for accuracy,
such as in-flight decay, stochastic emission of radiation.
In any case, the numerical integration of the equations of motion, which is what the
exercises are concerned with, requires stepwise raytracing through magnetic and/or
electrostatic fields defined in 3D space, possibly including time variation, in order
to ensure a faithful reproduction of the physical processes simulated.
Zgoubi [1] is resorted to here, however other well known codes use a numerical
integrator to push particles step-by-step in analytical or field map based field mod-
els, and could be used in various exercises, for instance RAYTRACE [2], GEANT [3],
to mention just two. Furthermore, some of the exercises involve simple optical
assemblies, for which a Runge-Kutta integration may easily be written.
Various exercises lend themselves to resolution using matrix or other mapping
transport techniques, however these techniques rely on approximations regarding
both field models and the solution of the equations of motion, resulting always on
questioning regarding the validity of one or the other, or both. The option of using
these techniques is left to the reader, however this is not our interest here, and no
such solutions to the exercises are provided: in the exercises, transport coefficients
. Ti j , Ti jk , ... and transport matrices .[Ti j ], [Ti jk ], ... will be addressed as a subproduct

© The Author(s) 2024 1

F. Méot, Understanding the Physics of Particle Accelerators, Particle Acceleration
and Detection, https://doi.org/10.1007/978-3-031-59979-8_1
2 1 Numerical Simulations

of stepwise raytracing, derived from particle coordinates, and mostly for comparison
with theoretical expectations regarding such quantities of paraxial beam optics as
magnification, optical aberrations, betatron functions, wave numbers.

1.1 Half a Century of Charged Particle Raytracing

in Zgoubi

At the time of this publication, 2024, the raytracing code zgoubi celebrates its 52st
anniversary, having pushed particles in accelerator laboratories over half a century.
The initial version of zgoubi was developed by D. Garetta and J. C. Faivre at
CEA-Saclay in the early 1970s, for the purpose of the design and operation of mag-
netic spectrometers at the 3 GeV polarized ion synchrotron SATURNE 2 [4, 5],
using magnetic field maps, simulated in zgoubi proper in the design approach
[1, AIMANT keyword] and measured eventually. The code was used to assess, from
their measured field maps, the optical properties of the magnets of the SATURNE 2
ring which was under construction in the same period.
The author of these lines inherited Zgoubi at SATURNE in the early 1980s for
spectrometer design and spin dynamics studies, from Saby Valéro who was com-
pleting the design of GANIL SPEG spectrometer [6], and André Tkatchenko [7].
The diversity of the utilization of zgoubi in the following years boosted the devel-
opment of analytical models of accelerator components, with today a library of
more than 60 optical elements and about 50 monitoring and command keywords
[1, cf., Glossary of Keywords].
In September 2007 it has been made available in sourceforge, https://sourceforge.
net/projects/zgoubi/ and has undergone a substantial number of downloads since,
from many countries as it appears (Fig. 1.1). The sourceforge package includes

Fig. 1.1 Zgoubi downloads in sourceforge [8], 4,500+ over the period 2007–2022, from 67
different countries from USA, China, France, Germany, India, etc., to Peru, Bangladesh, Vietnam
1.2 Raytracing with Zgoubi—Solving the Exercises 3

zgoubi, as well as an ancillary data treatment/graphic companion, zpop (some-

times forsaken though, in favor of gnuplot and awk tools). The sourceforge branch
https://sourceforge.net/p/zgoubi/code/HEAD/tree/branches/exemples/ offers about
300 simulations covering half a century of zgoubi usage:
– beam lines and spectrometers of all sorts,
– all possible ring accelerator methods,
– linacs, recirculating (RLA) and energy recovery linacs (ERL),
– effects of synchrotron radiation on beam and polarization,
– beam imaging using visible synchrotron radiation,
– collect and acceleration of short-lived beams,
– etc.

Most of these examples are drawn from real life R&D and teaching activities, and
as such have been the subject of laboratory technical notes and other publications.
It may be helpful to refer to the latter when undertaking simulations, many can be
found in the sourceforge branch https://sourceforge.net/p/zgoubi/code/HEAD/tree/
branches/publications/.
The first Users’ Guide dates back to 1988, at CEA-Saclay [9]. Accounting for fur-
ther developments, and in order to facilitate access to the program an English version
of the manual was written at TRIUMF in 1990 [10]. The code, which so far had been
used to push particles in beam lines and spectrometers, was introduced to the realm
of cyclic accelerators in the early 1980s, for the purpose of partial Siberian snake
design studies at SATURNE [11]. In the mid-1990s, the computation of synchrotron
radiation electromagnetic impulse and spectra was introduced to investigate, and
solved, synchrotron radiation interference issues at the LEP mini-wiggler beam pro-
file monitor. In the mean time, several new optical elements were added, including
all sorts of electro-magnetic and electrostatic bends and lenses. Zgoubi has under-
gone extensive developments in the recent years, for design and machine operation
studies regarding high energy accelerators and storage rings, including the Neutrino
Factory FFAG rings in the early 2000s, polarized beams at the Brookhaven National
Laboratory 25 GeV AGS, its 3 GeV Booster, RHIC heavy ion collider, and the elec-
tron rings of the Electron-Ion Collider (EIC) complex at BNL. All this reflects in the
simulation exercises proposed in this book.

1.2 Raytracing with Zgoubi—Solving the Exercises

Zgoubi is a stand alone series of Fortran files, compiling does not require any spe-
cific library. Running zgoubi requires no interface (various interfaces have been
developed over the years though, and made available, see Sect. 1.3).
A beam optics problem in zgoubi consists in an ASCII input data file, its default
name is zgoubi.dat. That ASCII file may actually be split, in as many ancillary files
as desired, for instance according to a modular structure of an optical sequence.
4 1 Numerical Simulations

Executing zgoubi.dat is as simple as this: [pathTo]/zgoubi-code/zgoubi/zgoubi i.e.

typing the address of the executable file. The execution produces an output ASCII list-
ing, zgoubi.res, always. Zgoubi may produce various additional output files during
execution, according to user’s requests.
One has to bear in mind that the only thing zgoubi knows to do is pushing par-
ticles: starting from an initial position and velocity, it computes particle coordinates
along an optical sequence, by stepwise integration of the Lorentz force differential
equations of motion. The input data file describes that optical sequence; it also
includes diverse commands aimed at delivering ancillary results, the latter anyway
derived from particle coordinates. As aforementioned a few things may actually hap-
pen while particles are pushed: spin motion, decay in flight, synchrotron radiation,
space charge perturbation, etc.
An optical sequence in zgoubi is a sequence of keywords, most of them followed
by one or more lines of numerical data (e.g., in the case of optical elements: length,
field value, integration step size, fringe field parameters possibly, etc.), like so:

Title: this is my optical sequence. Particles will be

! pushed through, all the way to ’END’
’OBJET’
a few lines of data define initial particle coordinates (initial
conditions are needed to solve the differential equation of motion!)
’DIPOLE’
a few lines of parameters: field, fringe fields, etc.
! this is a comment line
’FAISCEAU’ ! print out local particle coordinates
’QUADRUPO’
a few lines of parameters: field, fringe fields, etc.
’DIPOLE’
a second dipole
an empty line, not a problem
’BEND’
another type of dipole, with its own parameters and subtleties
’DRIFT’
drift length
’FAISCEAU’ ! print out local particle coordinates
’SYSTEM’
2 ! 2 commands follow
echo ’this is a system call’
gnuplot < ./gnuplot_ellipses.gnu ! some gnuplot script
’END’ ! execution stops here
trash ! whatever follows is trash, ignored
more trash

An optical sequence begins with a title line. And then:

1.2 Raytracing with Zgoubi—Solving the Exercises 5

OBJET: most of the time the first keyword, it defines the coordinates of particles
making up the object to be transported; this is mandatory as initial conditions are
needed in order to solve the Lorentz force equation.
Optical elements and commands follow, for instance

– DIPOLE: define a dipole magnet;

– EBMULT: a combined E, B multipole;
– ELCYLDEF: a cylindrical electrostatic deflector; MULTIPOL: lenses; CAVITE
to accelerate; TOSCA to handle field maps; WIENFILTER; etc.

Zgoubi offers a total of about 50 magnetic and/or electrostatic optical elements

[1, pp. 9, 10 and 13, 14].
Commands—which are keywords as well—are added wherever desired along the
optical sequence, they include such procedures as

– FAISCEAU, FAISTORE: log local particle coordinates, respectively in zgoubi.res

or in an ancillary output file;
– IMAGE[S]: compute local image density and size, etc.;
– GOTO: move the execution pointer to some arbitrary location along the sequence
(useful for instance for managing beam transport amongst recirculating linacs
spreader and combiner sections);
– TWISS, MATRIX: compute paraxial quantities from rays; SYSTEM: a system
call;
– INCLUDE: to include ancillary input data files, a recursive command.

Keywords include switches, for instance to request

– spin tracking: SPNTRK, whose numerical data include initial spins, a necessary
ingredient as initial conditions are needed in order to solve the Thomas-BMT
equation;
– space charge perturbations: SPACECHARGE;
– in-flight decay: MCDESINT, synchrotron radiation: SRLOSS, etc.

Launching matching procedures resorts to FIT, FIT2 keywords, two different

matching methods.
In the exercises, optical elements and procedures are most of the time referred to by
their corresponding keyword, with little additional explanation: further information
regarding their use and functioning is to be found in the indispensable companion
to the resolution of the exercises, Zgoubi Users’ Guide [1]:

– PART A of the guide describes what keywords do and how, and the physics content
of the code, optical elements in particular.
– PART B details the formatting of the input data which follow most keywords (a few
keywords do not require any data, for instance YMY, FAISCEAU, MARKER).
– A complete list of the available keywords can be found in the “Glossary of
Keywords” sections at the beginning of both PART A and PART B.
6 1 Numerical Simulations

– A quick overview of what optical elements can be simulated using zgoubi,

and what keywords can be used for that, is given in the “Optical elements ver-
sus keywords” sections which follow the “Glossary of Keywords” sections. Note
in passing, there are most of the time various ways to simulate one particular optical
element, either for historical reasons, or to allow for actual and/or real life sub-
tleties (for instance, between a gradient dipole and an offset quadrupole; between
the various modes of operation of an accelerating radio-frequency system).
– The Index at the end of Zgoubi Users’ Guide is a convenient tool to navigate
keywords.

A concise notation KEYWORD[ARGUMENT1, OPTION, ...] is used in the exer-

cises and solutions: it is believed that the reader will get promptly familiarized with
these shortcuts, of which the main goal is to alleviate the text. The nomenclature
KEYWORD[ARGUMENT1, OPTION, ...] follows the nomenclature of the Users’
Guide, Part B. Three examples:

– OBJET[KOBJ = 1] stands for keyword OBJET (generating particle coordinates),

and KOBJ = 1 option retained here;
– DIPOLE[IL = 2, XPAS = 2.5] stands for keyword DIPOLE (magnetic dipole);
print out stepwise particle data to zgoubi.plt (this is what “IL = 2” stands for!);
integration step size XPAS = 2.5 cm;
– OPTIONS[CONSTY ON, WRITE OFF] stands for keyword OPTIONS (gives
access to various options), and two options retained here, (i) CONSTY (main-
tain constant transverse coordinates during stepwise integration through optical
elements), switched ON; (ii) switch off most print outs to zgoubi.res.
– INCLUDE[NBF = N,FNAME = fileName, LBL_1A = from_A,LBL_1B = to_B]
inserts locally, N times, a piece of a sequence copied from ‘fileName’ file,
comprised between LABEL1-type MARKERS ‘from_A’ and ‘to_B’.

Coordinate nomenclature
In the theoretical reminders, i.e. Sect. 1.3 in the following chapters, conventional
notations are used for particle coordinates, namely,

radial axial

. x, x , y, y ' , δs, δp/ p

'

transverse coordinates longitudinal

δp and .δs are respectively the momentum and path length offsets compared to a ref-
.

erence particle. These coordinates are defined in the Serret-Frénet frame, or moving
frame, Fig. 1.2.
In the exercises instead, zgoubi coordinates are used, namely

radial axial

. Y, T , Z , P , S, D
transverse coordinates longitudinal
1.2 Raytracing with Zgoubi—Solving the Exercises 7

y
refe
ren
M
ce

x
v
M0
s

Fig. 1.2 Moving frame .(M0 ; s, x, y) along a reference line. . M0 , at path distance .s from some
origin, is the reference particle location

ry
e cto
←
V Traj

Z
P
M

Y
←
W
T
0
Refer X
enc e

Fig. 1.3 Coordinates.Y ,.T ,. Z ,. P in zgoubi [1, Sect. 1.1]. Reference curve: a straight axis in optical
elements defined in a Cartesian frame; an arc of a circle in those defined in a cylindrical frame.
. O X : in the direction of motion, tangent to the reference; . OY : normal to . O X ; . O Z : orthogonal to
the .(X, Y ) plane; .W: projection of the velocity, .v, in the .(X, Y ) plane; .T : angle between .W and the
. X -axis; . P: angle between .W and .v

The transverse coordinates are explicited in Fig. 1.3. . S is the path length, . D is the
relative rigidity of the particle, relative to a reference rigidity specified as part of the
initial object definition in zgoubi input data file. As a matter of fact, an initial object,
i.e. the set of initial coordinates of particles to be raytraced, and possibly their spins,
always has to be defined, for zgoubi to solve the differential equations of particle
and spin motion.
An important additional parameter is the integration step. Figure 1.4 displays the
position and velocity vectors of a particle in zgoubi frame, and a .∆s push from
position . M0 to position . M1 . That push is performed using a Taylor expansion in .∆s
[1, Sect. 1.2]. The integration step size is one of the available controls on the accuracy
of the integrator, when applied to the Lorentz force equation, or to the Thomas-BMT
8 1 Numerical Simulations

Fig. 1.4 Position vector .R )

and normalized velocity z u (M 1
vector (.u = v/v) of a particle M1
in zgoubi frame. A .∆s push )
M0
takes the particle from u( y
position . M0 to position . M1 M0
) x
Z
(M 1
R

0)
(M
Y

R
X
0
Reference

spin equation. It also controls the accuracy of the simulation of events, such as photon
emission, in-flight decay, etc.
Conventional and zgoubi coordinate notations may sometimes be used concur-
rently, for instance when equations from the main text are referred to, or resorted to,
in the exercises. This is presumably in contexts exempt of ambiguity.

Reference frames of optical elements

Optical elements in zgoubi define fields in a Cartesian reference frame: this is the
case for instance for MULTIPOL, BEND, EBMULT; or in a cylindrical reference
frame: case of e.g., DIPOLE, ELCYLDEF. And similarly for field map handling
keywords: CARTEMES, TOSCA[MOD.≤19], BREVOL use a Cartesian meshing,
whereas POLARMES, TOSCA[MOD.≥20] use polar or cylindrical meshing. Refer-
ring to Fig. 1.5: let a particle location M(X, Y, Z) project at m(X, Y) (the dashed
curve figures the projected trajectory). In the case of an optical element (figured as
a rectangular box) defined in Cartesian coordinates, X and Y in zgoubi.plt (columns
respectively 22 and 10 [1, Sect. 8.3]) denote the coordinates taken along the fixed

Cylindrical Frame
m
Y
X
Cartesian Frame
RM

m
Y
AT
O X
O

Fig. 1.5 Cartesian and cylindrical reference frames in optical elements

1.3 Graphics, Data Treatment: zpop, gnuplot, awk, python 9

reference frame axes. In the case of an optical element (figured as an angular sec-
tor AT with some reference radius RM) defined in a cylindrical coordinate frame
.(Y, X, Z ), Y is the radius, X is the polar angle, counted positive clockwise, Z is the
vertical coordinate (column 12 [1, Sect. 8.3]).

1.3 Graphics, Data Treatment: zpop, gnuplot, awk, python

An execution of a beam optics problem in zgoubi produces a listing, zgoubi.res,

always. However, when running a problem the user often requests logging of exe-
cution data in zgoubi.fai (produced by FAISTORE[FNAME = zgoubi.fai, or else])
and/or zgoubi.plt (produced as a result of IL = 2 flag, e.g. as in DIPOLE[IL = 2]).
The output file zgoubi.fai is a record of more than 50 particle data (coordinates,
spin, etc.) [1, Sect. 8.2], at the location(s) where the keyword is inserted in the optical
sequence.
The output file zgoubi.plt is a record of more than 50 particle data, step-by-step
(coordinates, fields, step size, etc,) [1, Sect. 8.3] while integration proceeds through
an optical element.
Beyond, a PRINT command available in several keywords allows specific print-
outs during raytracing. For instance, CAVITE[PRINT] will cause particle accel-
eration data to be logged in zgoubi.CAVITE.Out, which can then be accessed
from gnuplot scripts, to produce graphs, data treatment, or provide debugging
help. In the same line, one would get zgoubi.HISTO.out from HISTO[PRINT],
zgoubi.OPTICS.out from OPTICS[PRINT], zgoubi.PICKUP.out from PICK-
UPS[PRINT], zgoubi.SPNPRT.Out from SPNPRT[PRINT], etc. [1, Sect. 8].
Zpop [12], an old companion postprocessor of zgoubi’s, allows handling
zgoubi.fai and zgoubi.plt. It also allows brute reading of and plotting from any of
the other files mentioned above. Zpop is part of the sourceforge package, portable
on any linux and Mac OS. Quick to launch (in an xterm window), quick to operate.
After years of development and utilization zpop allows all sorts of graphs, and var-
ious post-processing, reading particle coordinates and other data from zgoubi.plt or
zgoubi.fai records.
Zpop menu 7, for instance allows plotting any variable entering the process of
pushing particles step by step and element by element, against any other. There are
of the order of 60 of them: particle coordinates, .E and .B field components, spin
components, RF phase, step size, optical element number, turn number, etc., as well
as derivatives or combinations [1, PART D, Sect. 1.3]. By experience, menu 7 answers
most of the needs of lattice studies and beam dynamics simulations.
Zpop menu 8, allows further treatment of data read from these output files from
a run, for instance drawing of synoptics with trajectories superimposed, Fourier
analysis of periodic motion, matching of Enge’s fringe field coefficients, etc.
Although this book is not a guide to the use of zpop, graphs found in the solutions
of simulation exercises often use the latter.
10 1 Numerical Simulations

When they are not produced using zpop, data analysis and graphic in the solutions
use gnuplot, an incredibly simple yet powerful tool, even more so when added awk
commands. By experience, gnuplot is quite suited as a graphic interface to zgoubi
output data files, awk adds a powerful data analysis and treatment tool, both combined
answer about any needs.
There is more, about python, following section.

1.4 Interface to Zgoubi?

Zgoubi can be run without an interfacing software, there is no need for that. Again,
all that is needed is (i) an input data file, zgoubi.dat, which starts with the definition
of initial coordinates, followed by a linear description of the optical sequence to be
raytraced through, and with a few commands sprinkled around, and (ii) the following
command:
.[pathTo]/zgoubi

which is the address of the executable. Execution results are logged in output files,
of which zgoubi.res a minima. Whatever is needed to handle the code is found in
Zgoubi Users’ Guide, which is part of the sourceforge package [1].

python [13]
A Zgoubi user quick startup has been written by beginners a few years ago [14].
This startup introduces to pyZgoubi, a python based interface to zgoubi developed
by Sam Tygier, which has its own web site [15] and at present maintained at RAL

and BNL. is an additional python interface, developed by a group from

Brussels university, available on internet as well [16].

Not strictly speaking python, but based on anyway, Sirepo accelerator simulation
package by Radiasoft company also offers an interface to zgoubi [17].

References

1. F. Méot, Zgoubi Users’ Guide. https://www.osti.gov/biblio/1062013-zgoubi-users-guide The

latest version of the guide, on Sourceforge: https://sourceforge.net/p/zgoubi/code/HEAD/tree/
trunk/guide/Zgoubi.pdf
2. S. Kowalski, H.A. Enge, RAYTRACE. Laboratory For Nuclear Science (MIT, Cambridge, MA,
USA, 1986). http://aea.web.psi.ch/Urs_Rohrer/MyFtp/RAYTRACE/raytrace1.pdf
3. S. Agostinelli et al., GEANT4—A simulation toolkit. NIM A 506(3), 250–303 (2003). https://
geant4.web.cern.ch/
4. J. Thirion, P. Birien, LE SPECTROMETRE II. (Rapport Interne DPhN/ME, CEA Saclay, 1975)
5. H. Catz, LE SPECTROMETRE SPES II (Rapport Interne DPhN/ME, CEA Saclay, 1980)
References 11

6. P. Birien, S. Valéro, Projet de spectromt̀re magnétique à haute résolution pour ions lourds. Note
CEA-N-2215 (CEA-Saclay, 1981)
7. A. Tkatchenko, F. Méot, Calculs optiques pour le spectromètre à kaons de GSI (Rapport Interne
CEA/LNS/GT/88-07, Saclay, 1988)
8. Zgoubi downloads on sourceforge: https://sourceforge.net/projects/zgoubi/files/stats/map?dates=2007-
09-01 to 2023-01-01&period=monthly
9. F. Méot, S. Valéro, Manuel d’utilisation de Zgoubi (Rapport IRF/LNS/88-13, CEA Saclay,
1988)
10. F. Méot, S. Valéro, in collaboration with J. Doornbos, P. Stewart, Zgoubi users’ guide, Int. Rep.
CEA/DSM/LNS/GT/90-05, CEA Saclay (1990) & TRIUMF report TRI/CD/90-02 (1990)
11. F. Méot, A numerical method for combined spin tracking and raytracing of charged particles.
NIM A313, 492, Proc. EPAC 1992, 747 (1992)
12. Zgoubi’s data treatment software zpop comes, and compiles independently, as part of the
zgoubi package [8], when downloaded from sourceforge. It is available, including source
files, at https://sourceforge.net/p/zgoubi/code/HEAD/tree/trunk/zpop/
13. https://www.python.org/
14. A. Pressman, K. Hock, Zgoubi. A Startup Guide for the Complete Beginner (2014). https://
sourceforge.net/p/zgoubi/code/HEAD/tree/trunk/guide/aGuide4Beginner/arxiv.org_abs_
1405.4921.pdf
15. S. Tygier, D. Kelliher, Developers: pyZgoubi. https://github.com/pyZgoubi/pyZgoubi
16. C. Hernalsteens, R. Tesse, M. Vanwelde, Zgoubidoo. https://ulb-metronu.github.io/zgoubidoo/
17. https://www.sirepo.com/en/apps/particle-accelerators/

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0
International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing,
adaptation, distribution and reproduction in any medium or format, as long as you give appropriate
credit to the original author(s) and the source, provide a link to the Creative Commons license and
indicate if changes were made.
The images or other third party material in this chapter are included in the chapter’s Creative
Commons license, unless indicated otherwise in a credit line to the material. If material is not
included in the chapter’s Creative Commons license and your intended use is not permitted by
statutory regulation or exceeds the permitted use, you will need to obtain permission directly from
the copyright holder.
Chapter 2
Electrostatic Systems

Abstract This chapter introduces to electrostatic systems used in beam optics, and
to the theoretical material needed for the simulation exercises. It begins with a brief
reminder of the historical and technological context, and continues with electrostatic
optics methods which beam handling, guiding and focusing lean on. Zgoubi opti-
cal element library offers analytical modeling of several electrostatic components.
For instance ELCYLDEF: an electrostatic deflector; ELMULT: a multipole, up to
20 poles; WIENFILTER: a plane condenser, possibly combining a magnetic dipole;
ELMIR, ELMIRC: N-electrode mirrors and condenser lenses, with straight or cir-
cular slits. Electrostatic elements can be simulated as well using field maps, via the
keywords TOSCA, MAP2D-E or ELREVOL. Running a simulation generates a vari-
ety of output files, including the execution listing zgoubi.res, always, and, on demand,
such files as zgoubi.plt, zgoubi.fai, zgoubi.MATRIX.out, aimed at looking up pro-
gram execution, storing data for post-treatment such as graphics, etc. Additional
keywords are introduced as needed in the exercises, such as the matching proce-
dures FIT[2]; FAISCEAU and FAISTORE to log local particle data in zgoubi.res or
in a user defined ancillary file; MARKER; the ‘system call’ command SYSTEM;
REBELOTE do-loop for multiple-pass or for parameter scans; and some more. This
chapter introduces in addition to spin motion in electrostatic fields, the simulation of
which is triggered by the keyword SPNTRK. SPNPRT or FAISTORE log spin vec-
tor components in respectively zgoubi.res or an ancillary file. The “IL = 2” flag logs
stepwise particle data, including spin vector, in zgoubi.plt file. Simulations include
deriving transport matrix, beam matrix, optical functions, from rays, using MATRIX
and TWISS keywords.

Notations Used in the Text

.A vector potential
.a electron gyromagnetic anomaly, .a = 1.15965 × 10−3
.B magnetic field
. Bρ magnetic rigidity, . Bρ = p/q

© The Author(s) 2024 13

F. Méot, Understanding the Physics of Particle Accelerators, Particle Acceleration
and Detection, https://doi.org/10.1007/978-3-031-59979-8_2
14 2 Electrostatic Systems

. E; .m 0 c2 ; . E i energy, . E = mc2 ; at rest; injection energy

.E; . E s,x,y ; . E ⊥ electric field vector; its components; normal component to .v
. Eρ electric rigidity, . Eρ/v = p/q = Bρ
.F Lorenz force
FOFDOD a Focusing-drift-Focusing-Defocusing-drift-Defocusing lattice
cell
. G hadron gyromagnetic anomaly. Proton: .G = 1.7928474
.m; .m 0 particle mass; at rest
.(O; r, θ, z) cylindrical frame
.(O; s, x, y) Cartesian frame
.p; . ps,x,y momentum vector of a particle; its components
.q particle charge
. R0 ; r 0 condenser equipotential radii
.s path variable [.(∗)' = d(∗)/ds]
.T kinetic energy
.t time variable [.(∗) ˙ = d(∗)/dt]
.U potential energy
.v; .vs,x,y velocity vector of a particle; its components
. V ; . Vi voltage ∫ ⊥ ds
.α trajectory deflection, .α = EEρ
.β .v/c

.δp/ p, .δ relative momentum offset

.φ electrostatic scalar potential
.ρ curvature radius

2.1 Introduction

A historical electrostatic beam line is the column of electrostatic tubes which, in

1932, allowed guiding and accelerating a proton beam to a target, so producing
the first artificial atom-splitting reaction, .p + 7 Li → 2 4 He, the Cockcroft-Walton
experiment [1]. A high voltage was produced by an ad hoc diode and condenser
cascade rectifying the AC voltage from a transformer. This high DC voltage was
applied to a string of conducting cylinders (Fig. 2.1) which ensured beam guiding,
(sufficient) focusing, and acceleration to 700 keV, a high enough energy to break the
Coulomb barrier in this nuclear reaction. Which earned its authors the 1951 Nobel
Prize.
Electrostatic systems allowed a landmark advancement, the first acceleration of a
polarized beam, at the University of Basel in the early 1960s, a period where polarized
proton and deuteron sources began operating [3, 4]. The experiment used a 200 keV
electrostatic accelerator. “The Basel group [...] presented the first deuteron source
in operation at the time of the first polarization conference in Basel 1960 ” [5]. The
convention for the sign of polarization is known as the “Basel Convention”. Polarized
beam acceleration at the nearby ETH Zurich 6 MV Van de Graaff generator was not
2.1 Introduction 15

Fig. 2.1 A similar tube cascade to the early 1930s Cockcroft-Walton experiment eponymous accel-
eration system: Fermilab’s 750 keV .H− injector [2]

far behind. Acceleration of polarized ions in cyclic accelerators soon followed, to

way higher energy, starting with cyclotrons [4].
A landmark in physics as well: the electron column. The design of the first elec-
tron microscope and of the scanning tunneling microscope earned their authors the
1986 Nobel Prize—well, actually these designs used magnetic lenses. Nevertheless,
the electron column, which combines electrostatic and magnetic components, is a
widespread system since, with a number of variants such as transmission-, scanning-,
or photoemission-electron microscope, and the electron-beam lithography column.
16 2 Electrostatic Systems

Fig. 2.2 Typical beam handling in an ion source region (BNL AGS injectors). Several electrostatic
systems are at work in a short distance: a focusing Einzel lens, a Wien filter mass selector, pre-
accelerating tubes, an inflector which serves as a switch with a Tandem ion line, electrostatic
condensers to steer the beam, more acceleration tubes

Electron beam energies range in 0.1–1 MeV [6]. A century of design and technolog-
ical refinements in electron optics have brought these systems to optical perfection.
Electrostatic optical elements are light objects. Deflectors and lenses are simple
to construct, simple mechanic forms shape the required fields, electrode voltages
can be up to a fraction of a MV, gradients to several MV/m, there is no remanence,
power consumption is low. All reasons why electrostatic optical elements are used
where energy allows, in low energy beam lines in particular (Fig. 2.2). Guiding and
focusing components include prism, plane condenser, multipoles, mirrors, etc. [12],
Figs. 2.3, 2.4, 2.5 and 2.6. Electrostatic components are not a specificity of low
energy lines though, they span a large range of applications, with energy and size
varying accordingly. On the small side are Einzel lenses used in particle source areas
(Fig. 2.3). Main bends in beam lines may be of larger volume (Fig. 2.4). Even larger,
in the meter range, are injection and extraction septa in GeV synchrotrons (Fig. 2.5),
or pretzel orbit separators in GeV e+ e– colliders such as LEP and CESR [13, 14]
(Fig. 2.6).
The electrostatic septum (Fig. 2.5) in particular is commonly used to steer beam
into or out of circular accelerators. Megavolts/m gradients allow handling high beam
rigidities, and achieve fraction of milliradian deflections aimed at. To give an idea
of quantities at stake, the septum in Fig. 2.5 for instance is an 80 cm long device,
2.1 Introduction 17

Fig. 2.3 Quite popular, the Einzel lens [7]. Three specimen here, diameters from 10 to 40 mm,
operation voltage 10 to 30 kV

Fig. 2.4 A 3-way spherical electrostatic deflector [9]. Beam can be switched left or right, or let go
straight
18 2 Electrostatic Systems

Fig. 2.5 A 250 kV septum for slow extraction at the SPS [8]. Electric field is on the extracted beam
side

Fig. 2.6 Cornell ESR 3 m long horizontal pretzel separator, operating voltage .±85 kV (2 MV/m).
Electrodes are split to let synchrotron radiation through
2.1 Introduction 19

Fig. 2.7 Elisa in Aarhus, a 25 keV, 7.6 m circumference racetrack for molecular and atomic
physics [10]. Its lattice combines spherical deflectors, plane deflectors and quadrupoles

Fig. 2.8 UMER ring at the University of Maryland [11]. A 10 keV, 11.5 m circumference beam
optics and beam dynamics test accelerator

septum thickness is 100. µm, operating voltage 260 kV (15 MV/m over a 17 mm gap)
for a deflection angle of 0.28 mrad.
Electrostatic optical elements have also invited themselves in the realm of rings.
An electrostatic ring is used every once in a while for proof-of-principle purposes.
The first occurrence was the “Electron Analog” (Fig. 2.9), built in 1954 to assess the
20 2 Electrostatic Systems

Fig. 2.9 The “Electron Analog”, a prof-of-principle of BNL AGS, a strong .n = 225 index FOF-
DOD lattice, 45 ft in diameter, built in 1954 [15]. The left column shows the cross sections of the
electrostatic optical elements which comprise the lattice

novel concept of strong focusing and transition-gamma crossing (cf. Chaps. 8 and
9), prior to the construction the AGS at the Brookhaven National Laboratory [15].
In the 1990s electrostatic rings were raised to the rank of tools for physics research,
with energies of keVs to tens of keVs. Examples are the ion storage ring ELISA
(Fig. 2.7), the beam physics ring UMER (Fig. 2.8), amongst others.
2.2 Basic Concepts and Formulæ 21

2.2 Basic Concepts and Formulæ

Mathematically speaking, electrostatic elements exploit the scalar potential compo-

nent in
∂A
.E = −gradφ −
∂t
allowing local deflection and/or focusing and/or acceleration along DC voltage gaps.
A fundamental aspect is that the resulting Lorentz force works. Particles exchange
energy with the field, at a rate .F · v = qE · v which is in general non-zero, thus
mass and velocity vary along the trajectory. This is a major difference with magnetic
elements, in which .F · v = q(v × B) · v ≡ 0, the magnetic force does not work, .|v|
and mass do not change.
Solving the Lorentz force differential equation . dmvdt
= qE requires the electric
field distribution in space. The latter derives from a potential solution of the Laplace
equation .∇ 2 φ = 0, The necessary boundary conditions to solve it depend on the
electrical properties of the device, on its shape, symmetries, and various components.
For instance electrodes are equipotentials to which the electric field is normal; the
electric field is along the axis in cylindrical tube; the transverse plane between two
identical iso-potential tubes is a symmetry plane, etc. In simple systems, or with
some ad hoc approximations, it is possible to find an analytical solution to the Laplace
equation, from which analytical expressions of the components of the field vector.E =
−gradφ may be derived. In some complicated cases, it may still be possible to find
analytical solutions for field components along a symmetry axis, or over a symmetry
plane, and extrapolate from there using Taylor expansion and Maxwell’s equations.
With complicated geometry the easiest way may end up being the computation of a
field map. Raytracing in a field map is at the expense of accuracy of the integration,
though, as a result of field interpolation from a mesh. A dense mesh and an integration
step size commensurate with the mesh size mitigate the issue.

2.2.1 Kinetics

Circular Motion; Rigidity

The Lorentz force on a particle of charge .q and mass .m in an electric field .E is

dp d(mv)
.F= = = qE (2.1)
dt dt
Circular motion requires velocity .v to be normal to the electric field .E. Deflectors
allow that, see below. It requires in addition, as in the cyclotron, the centripetal force
to equate .F. Write it under the form .q E = −mv 2 /ρ. Ignoring the sign, this yields
the electric rigidity
22 2 Electrostatic Systems

p2 T 1+γ m 0 c2 γ 2 − 1
. Eρ = = = (2.2)
qm q γ q γ

where.T = mc2 − m 0∫c2 is the kinetic energy of the particle. The trajectory deflection
over an arc of length . ds normally to the field is
∫ ∫ ∫
E ds 1 E ds 1 E ds
α=
. = = (2.3)
Eρ v p/q v (Bρ)

where .(Bρ) denotes the particle rigidity. The velocity .v appears in the expression for
the deflection angle, compared to magnetic deflection .α = B L/Bρ.

Work of the Force

The work by a force .F in the time interval .t1 , .t2 , over .dM = vdt is
∫ t2
T1,2 =
. F(M, t) · dM (2.4)
t1

Developing yields
∫ ( ) ∫
t2
d m0v t2
m 0 v · dv
. T1,2 = · vdt =
t1 dt (1 − v 2 /c2 )1/2 t1 (1 − v 2 /c2 )3/2
∫ ( ) ∫
t2 t2
m 0 c2
. = d √ = d(mc2 ) = [m 2 − m 1 ]c2 (2.5)
t1 1− v 2 /c2 t1

Thus, with kinetic energy defined as .T = mc2 − m 0 c2 = E − m 0 c2 the work writes

T1,2 = E 2 − E 1 = T2 − T1
. (2.6)

If .F derives from a time-independent potential .V , namely F = −q gradV (M, t),

C
then, with .U = q V ,
∫ t2
T1,2 = E 2 − E 1 = T2 − T1 = −
. gradU · dM = U2 − U1 (2.7)
t1

thus
. E 1 + U1 = E 2 + U2 , T1 + U1 = T2 + U2 (2.8)
2.2 Basic Concepts and Formulæ 23

In the non-relativistic limit .v/c ≪ 1, .γ ≈ 1 + β 2 /2 so that, as expected

β→0 1
T1,2 = E 2 − E 1 −→
. m 0 (v22 − v12 ) (2.9)
2

Motion in a Uniform Field

Take the .x axis parallel to .E, .E = E x x. The equations of motion write

| dps |
| | ps = ps0
dp | dtx = 0 |
. = qE ⇒ || dp
dt
= q Ex thus | px = q E x t + px0
| (2.10)
dt | dp y = 0 | p y = p y0
dt

Simplify the developments by taking the motion parallel to the .s axis at time .t = 0,
|
| ps0
|
.p0 = | 0 (2.11)
|
|0

Integrating Eq. 2.10 is not straight forward as .m is a function of .v, such that
m 0 vs,x,y
. ps,x,y = /
vs2 +vx2 +v 2y
1− c2

The difficulty can be surmounted in two steps [16]:

(i) Take. E 2 = p 2 c2 + m 20 c4 , with. p 2 = ps2 + px2 + p 2y = ps0
2
+ (q E x t)2 , note. E(t =
0) = E i . Thus

. E 2 (t) = (m 0 c2 )2 + ps0
2 2
c + (q E x t)2 c2 = E i2 + (q E x t)2 c2 (2.12)

(ii) With .v = p/m = c2 p/E, and . ps0 = βi E i /c as .p(t = 0) = ps0 s, one then gets
|
| ds = vs = √ ps0 c2 = √ 2βi Ei c 2
| dt +(q E i +(q E x ct)
|
2 2
E i E x ct)
. | d x = vx = √
q E x c2 t (2.13)
| dt
| E i2 +(q E x ct)2
| dy = v = 0
dt y

An interesting result here is that the longitudinal velocity decreases with time.
The transverse acceleration causes longitudinal deceleration. The radial velocity .vx
increases, with .c an upper limit:

dx q E x c2 t t→∞ q E x c2 t
. = vx = / −→ √ = ±c
dt E i2 + (q E x ct)2 (q E x ct)2
24 2 Electrostatic Systems

The trajectory slope increases linearly with time,

dx d x/dt q Ex q Ex c
. = = t= t (2.14)
ds ds/dt ps0 βi E i

Integrate the differential Eqs. 2.13:

|
| ds = √ ps0 c2 dt = ps0 c √ dt
, with a = Ei
| q E x a 2 +t 2 q Ex c
| E i2 +(q E x ct)2
. | d x = √ q E x c tdt
2
| = √ctdt (2.15)
| E i2 +(q E x ct)2 a 2 +t 2
| dy = 0

∫ ∫ √
On the one hand . √adt2 +t 2 = Asinh at ; on the other hand . √atdt 2 +t 2
= a 2 + t 2 , so
that
| ∫ [ ]
| s = ps0 c t √ dt2 2 = ps0 c Asinh t t = ps0 c Asinh q E x ct
| q E x 0 a +t q Ex a 0 q Ex [
/ Ei ]
| ∫ t tdt [√ ]t
.|x = c √ =c a +t 2 2 = q Ex
1
E i + (q E x ct) − E i
2 2 (2.16)
| 0 a 2 +t 2
| 0
| y = 0 (motion is in (O; s, x) plane)

The trajectory .x(s) is obtained by eliminating time between .x and .s using

q Ex s
q E x ct = E i sinh
. (2.17)
ps0 c

so that (accounting for .cosh2 − sinh2 = 1)

( ) ( )
Ei q Ex s Ei q Ex s
. x= cosh −1 = cosh −1 (2.18)
q Ex ps0 c q Ex βi E i

The motion is a catenary—the shape of a chain hanging by its two ends, under the
effect of gravitation (Fig. 2.11). A paraxial approximation, valid for a small enough
deflection, takes the Taylor development of .cosh, yielding a parabolic trajectory

1 q Ex 2 s2
x
. paraxial ≈ s ≈ (2.19)
2 βi2 E i 2ρ0

where .ρ0 = βi2 E i /q E x is the radius of the tangent circle to the parabola.
2.2 Basic Concepts and Formulæ 25

y
x

Ex

O
v s0
s

Fig. 2.10 Working frame (O; s, x, y). .E || x and .v(s = 0) || s

Zgoubi|Zpop Zgoubi|Zpop
22-12-2023 Y (m) vs. X (m) 22-12-2023 dp/p vs. s (m)

0.35
0.2
0.3

0.15 0.25

0.2
0.1
0.15

0.1
0.05
0.05

0.0 0.2 0.4 0.6 0.8 1. 0.0 0.2 0.4 0.6 0.8 1.

Fig. 2.11 Left: a catenary, trajectory of a 350 keV electron over 1 m in a . E s = 980 kV/m field.
Right: the evolution of its relative momentum offset .δp/ p0 from .s = 0 to .s = 1 m

2.2.2 Optical Components

As a particle travels in the electric field of an electrostatic element, its energy changes
because the field along the path is in general not normal to the velocity, .F · dM /= 0.
This affects the velocity and mass (Eqs. 2.4 and 2.5).
In optical elements a reference optical axis is defined, straight or curved depend-
ing on the device. The analytical formalism in general assumes paraxial optics, i.e.
trajectory angle to the optical axis remains small.
In various optical components, such as the Wien filter (see Sect. 12.2.4),
quadrupoles, toroidal deflectors, the electric field is normal to the optical axis. Impli-
cations are

– the field is considered normal to trajectories as well, longitudinal velocity is

assumed of constant magnitude,
– transverse excursions are small so that energy change can be ignored.
26 2 Electrostatic Systems

Fig. 2.12 A sketch of a circulating beam

plane condenser used as an
electrostatic septum, a septum
separation between a
circulating beam, unaffected, extrac
t
and an extracted beam
E beam ed
deflected under the effect of
the field
electrode

Things are different in cylindrical lenses and mirrors, where the electric field can
be near parallel, or far from normal, to trajectories.
These assumptions, aimed at allowing simplifying hypotheses for the sake of
analytical modeling, are irrelevant anyway if numerical integration is used to solve
the Lorentz equation.

Transverse Fields

Plane Condenser
A plane condenser is a simple concept (Fig. 2.12): a pair of parallel plates, to which
a voltage is applied, allowing the deflection of a charged particle beam. The device is
used in various optical systems: for beam guiding in low energy beam lines, electron
columns and ion rings; for beam switching; in accelerators up to high rigidities for
peeling out or switching beams; for orbit separation in high energy e+ e– colliders,
to mention a few.
The paraxial approximation of the deflection .α ≈ tan α undergone over a distance
. L in the uniform field can be obtained from .tan α = d x/ds (Eq. 2.14), using Eq. 2.17

to remove time, giving

q Ex L q Ex L
.α = = 2 (2.20)
βi ps0 c βi E i

At this point it is interesting to compare with the equivalent effect of a force of

magnetic origin, writing .q E = qcβ B. Thus,

. E = cβ B or E [GV/m] ≈ 0.3β B[T]

A deflection equivalent to that due to a 1 T magnetic field could be achieved with an

electric field of 9 MV/m in the case of a .β = 0.01 particle, but is not doable for a
.β ≈ 1 particle.

In the paraxial, parabolic approximation (Eq. 2.19), the radial motion writes [9]
[ ] 2
δp s
. x(s) = x0 + x0' s + (2 − β ) − 1
2
(2.21)
p 2ρ0
2.2 Basic Concepts and Formulæ 27

Fig. 2.13 A sketch of an R0

electrostatic bend. A region
of radial electric field is
defined between concentric reference
r0
electrodes (equipotential
surfaces) with axial and
radial symmetry. The
curvature radius of the R
reference trajectory .ρ0 = r0

with .s the longitudinal coordinate in the condenser frame (Fig. 2.10).

The length of the trajectory from the origin at (X = 0,Y = 0) (assuming .v ⊥ E at
that location), to (X, Y(X)) along the condenser is (Eq. 2.18)
∫ X[ ]1/2
. l(X ) = 0 1 + Y '2 (X ) d X (2.22)
[∮ (∮ )2 ]1/2
∫X
= 0 1 + βi sinh a
1 X
d X = −ia Ei (i Xa , βi−2 )
1 X3 5
≈X+ 6β 2 a 2
+ ( 13 − 1
) 1 X
4β 2 10β 2 a 4
+ ···

with . Ei(x, k) the elliptic integral of the second kind, i the imaginary unit and, below,
a series approximation.
Section 12.2.4 further addresses the Wien filter, a combination of an electrostatic
plane condenser and a magnetic dipole, featuring .E and .B field components normal
to each other, and normal to the longitudinal axis.

Toroidal Condenser

A sketch of a toroidal condenser is given in Fig. 2.13, which also defines.r0 , the radius
of the reference axis and . R0 , the vertical curvature radius. The reference axis is in the
median plane, along an equipotential .φ = r0 E 0 /2 mid-way between the electrodes.
This class of electrostatic bend comprises

– spherical condensers, . R0 = r0 , electrostatic potential .φ = E 0 r ( 21 − ln rr0 ), and

– cylindrical condensers, .1/R0 = 0, , electrostatic potential .φ = Er ( rr0 − 21 ).

The deflection angle .α along the reference axis satisfies Eq. 2.3. The energy of
the ideal particle, along the optical axis, satisfies Eq. 2.2. Particle coordinates in a
moving frame (see Sect. 3.2.2, Fig. 3.8) can be defined, namely, .x = (r − r0 ) in the
bend plane, . y along an axis normal to the latter, and .s = r0 θ .
28 2 Electrostatic Systems

A .δp/ p off-momentum particle differs from the reference one by its mass and
velocity. The latter two vary as the particle travels across the bend, exchanging energy
with the field. Combine these effects, appropriate approximations lead to the linear
equations of motion in a cylindrical condenser (.1/R0 = 0) [9]

d2x 2 − β2 2 − β 2 δp d2 y
. + x = , =0 (2.23)
ds 2 ρ02 ρ0 p ds 2

By comparison with the equations of motion in a uniform magnetic field (see

Sect. 3.2.2, Eqs. 3.15 taken with a field index .k = 0), a factor .2 − β 2 appears, which
tends to .1 at relativistic energy, as .β → 1.
In a toroidal condenser (.r0 /R0 /= 0), in the non-relativistic case (.β ≈ 0), the equa-
tions of motion write [12]

d2x 2−c 2 δp d2 y c r0
. + x= , + 2 y = 0, with c = (2.24)
ds 2 ρ02 ρ0 p ds 2 ρ0 R0

Quadrupole

With the force parallel to the electric field, transverse focusing requires (in an (x,y)
plane transverse to the quadrupole axis)

∂φ ∂φ
. E x = −K x = − , E y = +K y = − (2.25)
∂x ∂y

A ‘–’ sign for . E x is a convention. Thus .E derives from the scalar potential

K 2
. φ= (x − y 2 ) (2.26)
2
In the case of a potential .±V /2 applied on the electrodes (Fig. 2.14), with radius .a
at pole tip, then . K = −V /a 2 .
The equation of the equipotential is
/
2φ
. y = ± x2 − (2.27)
K

an hyperbola with its axes at .45 deg to the coordinate axes. As a matter of fact, pause
( ) ( )( ) ( ) ( )( )
u cos 45◦ − sin 45◦ x x cos 45◦ sin 45◦ u
= , so that =
sin 45◦ cos 45◦ − sin 45◦ cos 45◦
.
v y y v

In this change of axes, .φ changes to .φ ∗ = K uv. Thus, an electrostatic quadrupole

skewed by .45 deg achieves the same focusing as a magnetic quadrupole.
2.2 Basic Concepts and Formulæ 29

Fig. 2.14 An electrostatic quadrupole [18]. This one, a design for a 50 keV ion ring, operates in
the kVolt range

The equations of motion have the same form as in a magnetic quadrupole (see
Sect. 14.4.2), namely
[
d2 x −q V ±1
+ Kx x = 0 V
. ds 2
d2 y with K x = −K y = = (2.28)
ds 2
+ Ky y = 0 mv 2 a 2 |Eρ| a2
electric rigidity (Eq. 2.2)

Relative efficiency of an electrostatic quadrupole

From .F = q(E + v × B) it results an equivalence between . E and .βcB. Technology
does allow electric gradients beyond, say, 30 MV/m. For .β = 0.1 for instance, the
effect of such electric field is equivalent to that of . B = 30 × 106 /0.1c = 1 T; for
.β = 1 it corresponds to . B = 30 × 10 /c = 0.1 T. This relative inefficiency limits
6

the use of electrostatic lenses to low energy beam lines.

Optical aberrations

More on the electric quadrupole can be found in [17]. Simulation outcomes reported
in the latter demonstrate the accuracy which numerical integration allows on high
order optical aberrations.
That publication also addresses the cancellation of second order achromatic aber-
rations by a lumped .(E, B) quadrupole. The use of a pair of those as the final focus
quadrupole doublet in a nanoprobe is the subject of Sect. 12.2.2.
30 2 Electrostatic Systems

Electrostatic Mirrors

Plane condensers include electrostatic mirrors [19]. These devices can be used for
great trajectory deflection, including mirroring. In the latter case the longitudinal
component of the velocity cancels and changes sign, a motion which stepwise ray-
tracing handles efficiently.
Sketches of two such devices, available in Zgoubi optical element library, are
given in Fig. 2.15.
The potential in the straight slit mirror can be modeled by (after [19], using the
notations of Fig. 2.15)

∑N
Vi − Vi−1 sinh(π(Z − Z i−1 )/D)
. V (Z , Y ) = arctan (2.29)
i=2
π cos(π Y/D)

where . N is the number of plate pairs, D is their gap. This model assumes mid-plane
symmetry, and infinite slits of negligible width. The mid-plane field component . E Z
(and derivatives if needed) is obtained by differenciation, . E Z = −d V (Z )/d Z .
The potential of the circular slit mirror can be modeled by

∑N ( )
Vi − Vi−1 π(r − Ri−1 )
. V (r ) = arctan sinh (2.30)
i=2
π D

This model assumes mid-plane symmetry, and slits of negligible width. The mid-
plane field . Er (r ) (and its r-derivatives if needed) are first derived by differenciation.
.E(r, Y ) is then obtained by Taylor expansion in .Y , using symmetries and Maxwell

relations [20].
An example of a design of a time-of-flight ring for mass separation, based on
these optical elements, is displayed in Fig. 2.16. More on this device can be found
in [21].

Cylindrical Lenses

Cylindrical lenses are used for their focusing properties, in some cases combined
with longitudinal acceleration. Focusing stems from the change of radial velocity
through the gap between the tubes. It can be written
∫
q Er (r, z)
∆vr =
. dz,
(gap) mvz

with .z the longitudinal axis, .r the radial coordinate, and assuming revolution sym-
metry.
 Traje R
ctory
2.2 Basic Concepts and Formulæ

x X
r
u Traje u r
s ctory
w
Symmetry
plane

Z w R1
O
L1 L2 L3
R2
r
u
V1 V2 V3
V1 V2 V3
Y
Y
Mid−plane
D
D
Z

Fig. 2.15 Electrostatic N-electrode lens condensers (N = 3 electrodes, here), with straight slits on the left, circular slits on the right. They are used as deflectors
in this schematic, however the device can be used as well in mirror mode, or as a focusing or defocusing lens in transmission mode. .(Z , X, Y = 0) is the bend
plane. Parameters which define these systems include voltages .V1−3 , plate lengths . L 1−3 and slit locations . Z 1,2 or . R1,2
31
32 2 Electrostatic Systems

Fig. 2.16 A low energy electrostatic storage ring employed as a multiturn time-of-flight mass
spectrometer. Three-plate condensers are used for both focusing (LH1-4, horizontal and LV1-4,
vertical) and bending (M1A-B and M2A-B)

V1 V2 V1
R0

L1 D L2 D L3

Fig. 2.17 Unipotential lens, with revolution symmetry. The tubes are distant. D, they have the same
diameter, .2R0 . Their lengths and potentials are respectively . L 1 , V1 , . L 2 , V2 and . L 3 , V1

Numerical integration of the Lorentz equation along a trajectory requires knowing

the potential. The electric field results, which provides the force which applies on
the charged particle. Numerous publications have been dealing with the analytical
modeling of cylindrical lenses, and testing of these models. Below are two examples,
found in zgoubi optical element library.

Unipotential Lens [17]

A schematic of an unipotential lens is given in Fig. 2.17. Revolution symmetry about
the .z axis is assumed here. Various models for the electrostatic potential along the
axis can be found in the literature, not so different anyway. A possibility is [22]
⎡ ⎤
ω (z + 1/2L 2 + D) ω (z − 1/2L 2 − D)
cosh cosh
V2 − V1 ⎢⎢ln R0 R0 ⎥
⎥
. V (z) = + ln
2ωD ⎣ ω (z + 1/2L 2 ) ω (z − 1/2L 2 ) ⎦
cosh cosh
R0 R0
(2.31)
The origin for .z is taken in the middle of the central electrode, and .ω = 1.318.
Differenciation provides the electrostatic field component along the longitudinal
axis, . E z (z, r = 0) = −d V (z)/dz. Radial and azimuthal field components are null
2.2 Basic Concepts and Formulæ 33

Fig. 2.18 Bipotential lens.

The tubes have the same V1 V2
diameter, .2R0 , their lengths
are respectively . L 1 and . L 2 ,
R0 z
and potentials .V1 and .V2

L1 D L2

along the longitudinal axis. Their derivatives are not, off-axis Taylor expansions
provide .E(r, z) [20, Sect 1.3.2].
More on the unipotential lens can be found in [17], including raytracing outcomes
and comparisons with numerical solution of the differential equation of radial motion
.r (z), namely,
( | )2
d2 R 3 1 d V (z) ||
+ R=0 (2.32)
16 V dz |
.
dz 2 r=0

with . R(z) = r (z) V 1/4 (z) the Pitch variable.

Bipotential Lens

This is the basic optical block of a string of tubes, including multi-gap acceleration
columns. An analytical model for the potential along the axis in the geometry of
Fig. 2.18, in the case where the distance between the two tubes is negligible, is [23,
Chap. 5, Sect. 5.1.2] [20, cf. EL2TUB]

V2 − V1 ωz V1 + V2
. V (z) = tanh + (D = 0) (2.33)
2 R0 2

such that . E z (z, r = 0) = −d V (z)/dz. The origin for .z is half-way between the
electrodes, and .ω = 1.318.
A second model assumes that the distance. D between the two tubes is large enough
that the field fall-offs from the two lenses do not overlap. It is written

z+D
cosh ω
V2 − V1 1 R0 V1 + V2
. V (z) = ln + (D > R0 ) (2.34)
2 2ωD/R0 z−D 2
cosh ω
R0

Figure 2.19 shows examples of converging proton trajectories computed using that
model.
If a string of more than 2 tubes is modeled, an accelerating column for instance,
an upstream end lens (respectively downstream end) is modeled using .V1 = 0 (resp.
. V2 = 0).
34 2 Electrostatic Systems

Zgoubi|Zpop
27-01-2024 Y, Z (mm) vs. s (m)

0.0
V1 V2

-1

-2
0 2 4 6 8 10

Fig. 2.19 Converging trajectories of 20 keV protons, with bipotential lens geometrical parameters
= L 2 = 36 cm, . R0 = 10 cm, . D = 1.2 R0 , and gap voltage .V2 − V1 = −13 kV
.L 1

2.2.3 Periodic Structures

Periodic electrostatic structures are typically found in rings. ELISA, UMER and the
Electron Analog are three examples, respectively Figs. 2.7, 2.8 and 2.9.
In the aforementioned hypotheses of paraxial optics, electric fields normal to
the velocity vector, assuming negligible energy exchange between the beam and
the electric field, particle motion abides by the principles of betatron motion. Basic
theoretical material can be found in Chaps. 3–9.
These assumptions may however be misleading, acceleration or deceleration in
the course of betatron motion may have noticeable effects. Fringe fields may also
affect particle motion. This in addition translates into coupling between transverse
and longitudinal motions. Stepwise raytracing is exempt of these limitations, as field
models can be made as accurate as necessary, whereas numerical integration accounts
for possible energy variation and coupling.
Electrostatic rings are typically synchrotron style of beam instruments. Longitu-
dinal beam handling can use RF systems, for beam bunching, or for acceleration or
deceleration. Bend and lens voltages are ramped during acceleration. Note that the
latter may in principle be faster than with magnetic optics where eddy currents are a
restricting factor.

Cyclic Acceleration Using an Electrostatic Field?

Is it possible to accelerate on a closed orbit using a DC voltage? The answer is ’no’.
The work of the force .F = qE over a path from A to B (top Fig. 2.20) only
depends on the initial and final states, .U A = q V A and .U B = q VB (Eq. 2.7), it does
not dependent on the details of the path. Thus, using an electrostatic field (.E =
−gradV (R)) it is not possible to accelerate a particle traveling on a closed path
2.2 Basic Concepts and Formulæ 35

Fig. 2.20 Top: the work of

the electrostatic force only
depends on voltages at A and
B, .V A and .VB , independently B
of the path. Bottom: case of a
closed path, the particle loses
A
along (2) the energy gained
along (1) (1)
A B

(2)

∮
(bottom Fig. 2.20) as . F.ds = 0. As a consequence, a DC voltage gap in a circular
machine does not produce energy gain.

2.2.4 Spin Precession

Consider the classical model which, to the spin angular momentum .S of a particle of
charge.q and mass.m, associates the magnetic moment.μ = (1 + G) 2m q
S of a spinning
charge [24, Sect. 2.2]. In that model, under the effect of an ambient magnetic field
.Ba , .S undergoes a torque

dS q
. = (1 + G) S × Ba (2.35)
dt 2m
A particle traveling in the electrostatic field .E of an optical system experiences in its
rest frame a magnetic field (.Ba ) which is the Lorentz transform of the former (it also
experiences a electric field, which does not couple to .μ). Expressing .Ba in terms of
the laboratory electric field yields the differential equation of spin precession,

dS ω
. =S× (2.36)
ds Bρ

around a precession vector

( )
1 E×β
.ω = γ a+ (2.37)
1+γ c

In these expressions, .S is in the particle frame, it has not been Lorentz-transformed,

and all other quantities, including time, are expressed in a laboratory frame.
36 2 Electrostatic Systems

2.3 Exercises

2.1 Plane Condenser; Spin Motion

Solution 2.1
Electron dynamics in a parallel plate condenser is considered in this exercise. Use
WIENFILTER to simulate it, hard-edge field model.
Take condenser length 1 m, and electric field .|E| = 0.98 MV/m. Note: the reason
for this electric field value is to be found in Exercise 12.13 et seqs., an optimal field
setting of a Wien filter used as a spin rotator.

(a) Produce a graph of a symmetric catenary across the condenser.

Check the transverse excursion of a particle and trajectory length, versus theory.
(b) Produce a graph of .(Ynum − Yth )/Yth as a function of integration step size. .Ynum :
particle excursion at the downstream end of the condenser, from numerical inte-
gration. .Yth : theoretical expectation, Eq. 2.18.
Use REBELOTE[IOPT=1] as a do-loop, changing the integration step size,
WIENFILTER[XPAS], at each pass.
(c) Add spin, parallel to the electric field .E at start. In a similar manner to (b)
produce a graph of .(θnum − θth )/θth as a function of integration step size. .θnum :
spin angle at the downstream end of the condenser, from numerical integration.
.θth : theoretical expectation, Eqs. 2.36 and 2.37.

For spin tracking, add SPNTRK.

2.2 Toroidal Condenser

Solution 2.2
Use ELCYLDEF to simulate a toroidal condenser.
(a) Set up a simulation showing that, in the paraxial hypothesis, in a √ cylindrical
condenser, a diverging beam is re-focused after a deflection .α = π/ 2.
Test the convergence of the numerical solution versus integration step size.
(b) Produce the aberration curve .T (Y ) at the focal plane. The moving frame can be
shifted to the latter using AUTOREF.

2.3 A Time-of-Flight Mass Spectrometer Ring

Solution 2.3
Multiturn storage is a convenient way to achieve high resolution mass separation, in
a compact apparatus. Electrostatic mirrors are potential candidates as both deflector
and focusing optical element for a low energy storage ring. The design displayed in
Fig. 2.16 is an example. This exercise reviews various of its aspects.
(a) The parameters of the ring are given in Table 2.1. Use ELMIR with appropriate
MOD option for both focusing lenses (LH, MOD = 22 and LV, MOD = 21) and
bends (MA and MB, MOD = 11).
2.3 Exercises 37

Table 2.1 Time-of-flight mass spectrometer. The parameters of a half-cell of the ring are given, as
the cell is symmetric. Referring to Fig. 2.16: this parameter list starts from the center of the long
drift (.s = 0), going clockwise
Particle .N2

Mass uma 28
Mass GeV 26.082
Charge |e| 1
Kinetic energy keV 400
Geometry
Ring circumference.a cm 393.73658
Gap height in condensers m 0.012
Number of slits in LH, LV 2
Number of slits in M 6
Length, electrode lengths
Drift (cm) 30.7
LV1 (3 .× cm) 2.525, 1.25, 2.525
Drift 1.2
LH1 (3 .× cm) 2.525, 1.25, 2.525
Drift (cm) 11.6
M1 (7 .×cm) 4.275, 5 .× 0.4163, 10
Drift (cm) 6.00217933
Electrode voltages, in that order
LV1 (V) 0, 115, 0
LH1 (V) 0, 40, 0
M1A (V) 0, 5 .× 200, 400
a
. This is the length of the reference closed orbit

Build zgoubi input data file. Produce a synoptic of the ring in laboratory
coordinates.
Produce the ring tunes, chromaticities. Produce a graph of the optical functions.
TWISS can be used for that.
Produce a graph of horizontal or vertical trajectory over a few tens of turns.
(b) Produce a chromaticity scan (i.e., wave numbers as a function of momentum
offset).
(c) Produce 1000-turn horizontal and vertical phase space motion, up to maximum
stable amplitudes.
(d) Produce the time-of-flight histograms after 20 turns, for a bunch comprised
of two masses: . M1 = 26.082 GeV and .1.0004 × M1. Both bunches have a
400 keV average energy, rms energy spread .δ E/E = 10−4 , rms emittances
−6
.ϵ x /π = 0.02138 10 m and .ϵz /π = 0.0106 10−6 m. All particles leave from
s = 0 at the same time.
Use PARTICUL[M=M1,M2] to define two different masses [20, cf. PARTICUL].
38 2 Electrostatic Systems

Table 2.2 Parameters of the AGS electron analog [15]

Injection energy, .Ti MeV 1
Maximum energy, . E max MeV 10
Physical radius, . R feet 22.5
Curvature radius, .ρ ft 15
Lattice cell FOFDOD
Number of cells, . N 40
Field index, .n 225
Phase advance per cell .≈ π/3

2.4 Converging Rays in a Bipotential Cylindrical Lens

Solution 2.4

(a) Reproduce Fig. 2.19, using the analytical modeling EL2TUB.

Check the evolution of proton rigidity across the lens.
(b) Repeat, using instead ELREVOL with a 1-D electrostatic field map.

2.5 The AGS Electron Analog

Solution 2.5
A schematic of the AGS electron analog is given in Fig. 2.9. Its parameters are given
in Table 2.2. Refer to Chaps. 8 and 9 for preliminary notions regarding betatron
motion.

(a) Based on these informations, build a simulation file of the electron analog.
Produce a graph of its optical functions.
(b) Accelerate an electron bunch, from 1 to 10 MeV. Produce a graph of the horizontal
and vertical phase spaces.

Check the betatron damping, compare with theory.

2.4 Solutions of Exercises of This Chapter: Electrostatic Systems 39

2.4 Solutions of Exercises of This Chapter: Electrostatic

Systems

2.1 Plane Condenser; Spin Motion

(a) A catenary
An input data file for a catenary is given in Table 2.3. The initial coordinates under
OBJET have been determined in the following way:
– perform a preliminary simulation in half the length, 50 cm, launching the electron
normal to .E. The expected transverse excursion is Y = 22.8948628 cm (Eq. 2.18),
this is what raytracing yields
– that simulation also gives the electron relative rigidity, D, after 50 cm, as well as
the trajectory angle, T.
The initial coordinates for the the 1 m long condenser are Y, –T and D.
The catenary and the evolution of electron momentum are displayed in Fig. 2.11.
Graphs obtained using zpop: menu 7; 1/1 to open zgoubi.plt; 2/[8, 2] for Y(X), or
2/[8, 1] for D(X); 7 to plot.

The final coordinates can be found in the execution listing zgoubi.res, under
DRIFT:

It can be verified that they are quite close to the initial ones, under OBJET in the
input data file, as expected.
This list includes the trajectory length, . S = 1.1275340 m, over . X : 0 → 1 m, or
half that value, .0.563767 m, over . X : 0 → 0.5 m. This is in agreement with Eq. 2.22
taken for . X = 0.5 m, .βi = 0.804837 and .a = βi E i /E x with . E i = 860998 eV and
. E x = 0.98 MV.

(b, c) Sensitivity to step size. Spin precession

An input data file for a scan of the step size is given in Table 2.4. The scan is based
on REBELOTE[IOPT = 1, LMNT = WIENFILT, KPRM = 8 0], which causes 1,000
times passes through the condenser, with each time the same initial conditions, and a
different value of the step size XPAS (parameter number 80 under WIENFILTER).
The step size value spans from 0.001 to 5 cm by .(5 − 0.001)/(1000 − 1) ≈ 5 ×
10−3 cm increments
The result of the scan is displayed in Fig. 2.21. Spin outcomes are found in
zgoubi.res, under “SPNTRK MATRIX”, namely (an excerpt):
40 2 Electrostatic Systems

Table 2.3 Simulation input data file: push an electron through a condenser, along a catenary

2.2 Toroidal Condenser

√
(a) Re-focusing at .α = π/ 2 rad √
The input data file to raytrace over.α = π/ 2 rad in a cylindrical condenser
√ is given in
Table 2.5. Figure 2.22 shows a pair of trajectories, re-focused after a.π/ 2 deflection.
Figure 2.23 shows the sensitivity of the result to integration step size. The graph
is unchanged whether .T0 = ±0.1 mrad or a hundred times less.
2.4 Solutions of Exercises of This Chapter: Electrostatic Systems 41

Table 2.4 Simulation input data file: push an electron through a condenser, along a catenary. Repeat
1,000 times with each time a different integration step size
42 2 Electrostatic Systems

0.1 1 10
-3 0
10*10 10*10
"zgoubi.fai" u ($0>2 ? stp_i + ($38-3)*dStep :1/0):(abs(($10-Yexp)/Yexp) )
"zgoubi.fai" u ($0>2? stp_i+($38-3)*dStep:1/0):(abs(atan($39/$38)-ttaexp)/ttaexp)

-3
1*10

0
1*10
|Y-Yth|/Yth

|θ-θth|/θth
-6
100*10

100*10-3

-6
10*10

1*10-6 10*10-3
0.001 0.01 0.1 1
step size [cm]

Fig. 2.21 Catenary motion. Sensitivity to step size for convergence of the numerical integration.
Left vertical axis, square markers: radial excursion. Right axis, circles: spin rotation

(b) Aberration curve at the focal plane

The previous data file can be used, with the following changes:
– use OBJET[KOBJ = 1, IMAX = 41] so to generate 41 particles launched with
. T0 ∈ [−20, 20] mrad, like so:

’OBJET’
2.33741826923 * 1.e3
1
1 41 1 1 1 1
.0 1. .0 0. 0. 1.
.0 0. .0 0. 0. 1.

– add AUTOREF[I = 2] after ELCYLDEF, that will cause the moving frame to
move to the waist formed by particles 1, 3 and 5 [20, cf. AUTOREF]
– add FAISTORE[FNAME = zgoubi.fai] after AUTOREF, to log particle data at
that location.
The following gnuplot script produces a graph of the horizontal phase space
(Fig. 2.24):

cm2m = 1e-2; mrd2rd = 1e-3

plot ’./zgoubi.fai’ u ($10 *cm2m):($11 *mrd2rd) w p ps .9 pt ; pause 2

2.3 A Time-of-Flight Mass Spectrometer Ring

(a) TOF ring data file. Optics
• Input data file. Synoptic.
.
2.4 Solutions of Exercises of This Chapter: Electrostatic Systems 43

Table 2.5 Input data file: track three positrons launched with initial horizontal angles .T0 =
0, ±0.1 mrad, through a cylindrical condenser

The input data file for this problem is given in Table 2.6.
A synoptic of the ring can be obtained using zpop, Fig. 2.25.

• Optical parameters, optical functions.

.

The input data file in Table 2.6 runs a TWISS command. This produces zgoubi.
TWISS.out, which contains the optical parameters. An excerpt:

@ LENGTH %le 3.311022574

@ ALFA %le 0.7808368683
@ ORBIT5 %le -0
@ GAMMATR %le 1.131670108
@ Q1 %le 0.25615960E+00 2 [frac., int.]
@ Q2 %le 0.28340770E+00 3 [frac., int.]
@ DQ1 %le 2.228316236
@ DQ2 %le 36.69163200
44 2 Electrostatic Systems

Zgoubi|Zpop
05-01-2024 Y (m) vs. α (rad)
95.56

95.54

95.52

95.5

95.48

95.46

95.44

0.0 0.5 1. 1.5

Fig. 2.22 Paraxial focusing in a cylindrical condenser. Launch angles√ of the three trajectories
shown are .0 and .±0.1 mrad. Location of convergence is after .α = π/ 2 = 2.22144 rad deflection

IMAGE offset vs. integration step size

0.35 0
"temp.gnu" u (stp0 + $0*dstp):(($11-X1)/X1)
"temp.gnu" u (stp0 + $0*dstp):(($15-Y1)/Y1)
-5
-2x10
0.3
-5
-4x10

0.25
-6x10-5
δX/X at image plane

δY/Y at image plane

-5
-8x10
0.2

-0.0001

0.15
-0.00012

0.1 -0.00014

-0.00016
0.05
-0.00018

0 -0.0002
0.3 0.32 0.34 0.36 0.38 0.4
step size [cm]

Fig. 2.23 Paraxial focusing in a cylindrical condenser. The figure shows the relative change in the
distance of the waist to trajectory 1 (which leaves OBJET with all coordinates null), longitudinally
(.δ X ) and radially (.δY ), as a function of integration step size

@ DXMAX %le 4.31815950E-01 -1.22847730E-08 @ DXMIN

@ DYMAX %le 4.69095543E-25 -4.82489222E-25 @ DYMIN
@ XCOMAX %le 1.18954794E+01 0.00000000E+00 @ XCOMIN
@ YCOMAX %le 0.00000000E+00 0.00000000E+00 @ YCOMIN
@ BETXMAX %le 1.02880410E+00 7.38003952E-01 @ BETXMIN
@ BETYMAX %le 1.30020403E+00 9.01549054E-02 @ BETYMIN

That file zgoubi.TWISS.out also contains the optical functions along the structure.
They are plotted using gnuplot_TWISS.gnu (Table 2.6), Fig. 2.26.
• Trajectories around the ring.
.
2.4 Solutions of Exercises of This Chapter: Electrostatic Systems 45

Horizontal phase-space, from zgoubi.fai.

0.0025
’./zgoubi.fai’ u ($10 *cm2m):($11 *mrd2rd):(i)

0.002

0.0015

0.001

0.0005
T [rad]

-0.0005

-0.001

-0.0015

-0.002

-0.0025
-5 -5
-0.0004 -0.00035 -0.0003 -0.00025 -0.0002 -0.00015 -0.0001 -5x10 0 5x10
Y [m]

Fig. 2.24 Aberration curve at the focal point of a 127.2 deg deflection toroidal condenser: a second
order (sextupole) aberration, .Y ∝ T 2 , typical of geometrical non-linearities in a bending element

Fig. 2.25 A synoptic of half 0.05

the TOF ring (except for the
0.0
central drift, not displayed),
and reference trajectories at LV MA
400 keV and .+/−20%. A
graph obtained using zpop: -0.1
menu 7, 2/[48, 42] for Y(X)
coordinates, 7 to plot
-0.2

LV MB
LH
-0.3

0.0 0.1 0.2 0.3 0.4

Trajectories along the structure are obtained using the previous input data file,
Table 2.6, with some changes, as follows. Use the following OBJET:
46 2 Electrostatic Systems

Table 2.6 Simulation input data file. Left: a half of the TOF ring, from middle of long straight. This
file also defines the sequence segment half-ring_S to half-ring_E, for INCLUDE in subsequent files.
Right: Computation of the optical functions using a TWISS, a SYSTEM commands runs a gnuplot
script taken from [pathTo]/zgoubi-code/toolbox/gnuplotFiles/gnuplot_TWISS/ folder, which plots
the optical functions read from zgoubi.TWISS.out

’OBJET’
15.23683 Rigidity (kG.cm) -> 0.4 KeV N2+ (mass=28)
8
1 1 1
0. 0. 0. 0. 0. 1.
-1.1897385E-004 7.3901428E-001 2.5e-99
-1.1658994E-001 1.0815484E-001 120e-9
0. 1. 0.

Particle data are logged in zgoubi.plt, this is triggered using OPTIONS:

2.4 Solutions of Exercises of This Chapter: Electrostatic Systems 47

Optical functions, from zgoubi.TWISS.out

1.4 0.45
βx βy ηx ηy
0.4
1.2
0.35
1 0.3
βx , βy [m]

0.8 0.25

η x , ηy
0.2
0.6 0.15
0.4 0.1
0.05
0.2
0
0 -0.05
0 0.5 1 1.5 2 2.5 3 3.5
s [m]

Fig. 2.26 Optical functions along the TOF ring, as read from zgoubi.TWISS.out

y (m)
1.0
E-4
(a) y (m) (b)
4.0
E-4
0.5
E-4 2.0
E-4

0.0 0.0

-2.0
-0.5 E-4
E-4
-4.0
E-4
-1.0
E-4 s (m) s (m)
0.0 0.5 1.0 1.5 2.0 2.5 3.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0

Fig. 2.27 Vertical trajectory of a particle over 80 turns around the ring. LH and LV potentials are
about .38 V and .76 V, respectively. a: initial .z(s = 0) = 0.07 mm. b: initial .z(s = 0) = 0.25 mm,
amplitude detuning brings the fractional tune near .ν y = 0.5, a island configuration in phase space
(see Fig. 2.29), causing the orbit closes in two turns. Graphs obtained using zpop: menu 7; 1/1 to
open zgoubi.plt; 2[6, 4] to select Z(S); 7 to plot

’OPTIONS’
1 1
.plt 200

Multiturn, 80 turns here, uses REBELOTE:

’REBELOTE’
80 0.1 99

Outcomes are displayed in Fig. 2.27.

48 2 Electrostatic Systems

Table 2.7 Simulation input data file: a tune scan, versus momentum. This file INCLUDEs the
sequence segment half-ring_S to half-ring_E defined in Table 2.6

Fig. 2.28 Chromaticities, 0.4

"zgoubi.MATRIX.out" every 1::4 u ($47-1):($45)
.dνx / dp/ p
"zgoubi.MATRIX.out" every 1::4 u ($47-1):($46)
(red, square 0.35
markers) and .dν y / dp/ p
(blue, circles) . LH and LV 0.3

voltages are 40 and

0.25
115 Volts, respectively
νx, ν y

0.2

0.15

0.1

0.05

0
-0.004 -0.002 0 0.002 0.004 0.006 0.008 0.01 0.012 0.014
δp/p

(b) Chromaticity scan.

An input data file for a scan of wave numbers as a function of momentum is given
in Table 2.7. Results are displayed in Fig. 2.28.

(c) 1000-turn phase spaces.

2.4 Solutions of Exercises of This Chapter: Electrostatic Systems 49

0.004
x’ (rad) (a) y’ (rad) (b)
0.003 0.001

0.002

0.001

0.0 0.0

-0.001

-0.002

-0.003 -0.001

x (m) y (m)
-0.004 -0.002 0.0 0.002 0.004 -0.001 -0.0005 0.0 0.0005 0.001

Fig. 2.29 Horizontal (a) and vertical (b) phase spaces, 1000 turns around the ring. LH and LV
potentials are about.38 V and.76 V respectively. Non-linear vertical motion causes amplitude detun-
ing towards fractional .ν y = 0.5 and island formation

Fig. 2.30 Histograms of the time of flight of masses M1 and M2, at 5 and 20 turns, observed at s
=0

Definition of initial trajectory coordinates can use OBJET[KOBJ = 2]. Use REBE-
LOTE[NPASS = 999, K = 99] for a 1000 turn loop. Outcomes are displayed in
Fig. 2.29.

(d) Histograms.
Definition of initial trajectory coordinates can use MCOBJET[KOBJ = 3]. Use
REBELOTE[NPASS = 19, K = 99] for a 20 turn loop. Outcomes are displayed in
Fig. 2.30.

2.4 Converging Rays in a Bipotential Cylindrical Lens

(a) Using the analytical modeling EL2TUB.
The input data file used to produce Fig. 2.19 is given in Table 2.8.

The graph is obtained using zpop: menu 7; 1/1 to open zgoubi.plt; 2/[6, 2] for .Y
versus .s or [6, 4] for . Z versus .s; 7 to plot
Proton rigidity is expected to increase from . Bρin = 0.020435 T m (20 keV) to
. Bρout = 0.026249 T m (33 keV) across the 13 kV gap voltage, a ratio of. Bρout /Bρin =
1.2845, confirmed in Fig. 2.31.
50 2 Electrostatic Systems

Table 2.8 Simulation input data file: tracking 20 keV protons in a bipotential cyclindrical lens

Fig. 2.31 Evolution of Zgoubi|Zpop

dp/p vs. s (m)
momentum offset 0.3 27-01-2024
.( p − p0 )/ p0 across the lens,
for all the protons tracked 0.25
(they all have .20 keV initial
energy). A graph obtained 0.2
using zpop: menu 7, 2/[6, 1]
0.15
for .( p − p0 )/ p0 versus .s; 7
to plot
0.1

0.05

0.0
0.0 0.5 1. 1.5 2. 2.5 3.
* # TRAJECTORIES - STORAGE FILE, 27-01-2024 17:40:02 *a
Mi-ma H/V: 0.00000E+00 3.00000E+00/ -2.00000E-02 3.00000E-01
Part# 1- 10000 (*); Lmnt# 1; pass# 1- 1, [ 1];
References 51

(b) Using ELREVOL with a 1-D electrostatic field map.

Proceed with the following steps:

• run the previous problem

– with a single particle, all initial coordinates and momentum offset null
– setting EL2TUB[IL = 2]

• read the step-by-step . X coordinate of the particle and the electric component
. E X (X ) so generated, across EL2TUB, from zgoubi.plt
• re-write that in an ascii file, say, el2tub.map, with proper format to be read by
ELREVOL (two columns, col. 1 is . X (step), col. 2 is . E X (step))
• change EL2TUB to ELREVOL with proper data list, with el2tub.map as field map
data file (refer to [20, Lookup INDEX for EL2TUB] for ELREVOL data list and
its formatting)

Run that new file. Essentially identical results are expected. Some discrepancy
may arise, reasons being

– the field map mesh size (i.e., the integration step size across EL2TUB): increase
mesh density as necessary for convergence of the trajectory results
– the integration step size across ELREVOL field map. It is mostly ineffective to
take it smaller than the mesh size.

2.5 The AGS Electric Analog

Building the AGS Electric Analog and operating it is left to the reader. All details
concerning the experiment and the ring can be found in [25, part D, Sect. 38].
ELCYLDEF can be used for the bends, WIENFILTER with null magnetic field
might work since the deflection is small (thus abiding by Eq. 2.20), to be confirmed.
ELMULT can be used for the two quadrupole families, focusing and defocusing.

References

1. J.D. Cockcroft, E.T.S. Walton, Experiments with High velocity positive ions. Proc. R. Soc.
Lond. A136, 619–630 (1932)
2. Figure 2.1: Credit Reider Hahn, Fermilab
3. T. Roser, A. Zelensky, Private Communication (BNL, 2021)
4. A.I. Yavin, The AVF cyclotron. Phys. Today 15(5), 19–25 (1962). https://doi.org/10.1063/1.
3058175
5. G. Clausnitzer, History of polarized ion source developments, in International Workshop on
Polarized Ion Sources and Polarized Gas Jets. ed. by Y. Mori (KEK, Tsukuba, Japan, 1990).
https://inis.iaea.org/collection/NCLCollectionStore/_Public/22/051/22051667.pdf
6. W. Wan, Aberration correction in microscopes, in TU4PBI02 Proceedings of PAC09 (Vancou-
ver, BC, Canada)
7. Figure 2.3: © Dreebit GmbH
52 2 Electrostatic Systems

8. M. Paraliev, in Proceedings of the CAS-CERN Accelerator School:Beam Injection, Extraction

and Transfer. ed. by B. Holzer (Erice, Italy, 2017); CERN Yellow Reports: School Proceedings,
Vol. 5/2018, CERN-2018-008-SP (CERN, Geneva, 2018), pp. 363–394, https://doi.org/10.
23730/CYRSP-2018-005 Figure 2.5: CERN, 2018. https://creativecommons.org/licenses/by/
4.0; no change to the material
9. P.J. Bryant, Transverse motion and electrostatic elements, Lecture 3, in Introduction to Particle
Accelerators. (Joint Universities Accelerator School, Archamps, 2010). Figure 2.4: copyrights
under license CC-BY-3.0, https://creativecommons.org/licenses/by/3.0; no change to the mate-
rial
10. S. Pape Møller, Design and first operation of the electrostatic storage ring, ELISA, in
Proceedings of EPAC’98 Accelerator Conference (Stockholm, 1998). https://accelconf.web.
cern.ch/e98/PAPERS/THZ01A.PDF Figure 2.7: copyrights under license CC-BY-3.0, https://
creativecommons.org/licenses/by/3.0; no change to the material
11. R.A. Kishek et al., Benchmarking space charge codes against UMER experiments, in
WEA3MP03 Proceedings ICAP 2006 (Chamonix, France, 2006). http://accelconf.web.cern.
ch/icap06/HTML/AUTHOR.HTM Figure 2.8: copyrights under license CC-BY-3.0, https://
creativecommons.org/licenses/by/3.0; no change to the material
12. A. Septier (ed.), Focusing of Charged Particles, vol. I, II (Academic, 1967)
13. W. Kalbreier, N. Garrel, R. Guinard, R.L. Keizer, K.H. Kissler, Layout, design, and construction
of the electrostatic separator system of the LEP e+ e– Collider, in Proceedings of EPAC, vol.
2 (1988) or CERN SPS/88-20 (ABT)
14. J.J. Welch et al., Commissioning and performance of low impedance electrostatic separators
for high luminosity at CESR, in Proceedings of the 1999 Particle Accelerator Conference (New
York, 1999). https://accelconf.web.cern.ch/p99/PAPERS/TUA156.PDF Figure 2.3: copyrights
under license CC-BY-3.0, https://creativecommons.org/licenses/by/3.0; no change to the mate-
rial
15. G.K. Green, E.E. Courant, The proton synchrotron. Part D, Sect. 38, The electron analog, in
Handbuch der Physik, Band XLIV (Springer, Berlin 1959), p. 319. Figure 2.9: Springer, All
rights reserved
16. G. Leleux, Accélérateurs Circulaires, in INSTN Lectures (Saturne, CEA Saclay, 1978). (unpub-
lished)
17. F. Méot, Generalization of the Zgoubi method for ray-tracing to include electric fields. NIM A
340, 594–604 (1994)
18. C.P. Welsch, Design studies of an electrostatic storage ring, in Proceedings of the 2003
Particle Accelerator Conference. Figure 2.14: copyrights under license CC-BY-3.0, https://
creativecommons.org/licenses/by/3.0; no change to the material
19. S.P. Karetskaya et al., Mirror-bank energy analyzers, in Advances in Electronics and Electron
Physics, vol. 89 (Academi, 1994), pp. 391–491
20. F. Méot, Zgoubi Users’ Guide. https://www.osti.gov/biblio/1062013-zgoubi-users-guide. An
up-to-date version of the guide can be found at: https://sourceforge.net/p/zgoubi/code/HEAD/
tree/trunk/guide/Zgoubi.pdf
21. M. Baril, F. Méot, D. Michaud, Design study of a compact multiturn time of flight mass
spectrometer. Internal Report CEA DSM DAPNIA/SEA-00-08 (2008)
22. A. Septier, Cours du DEA de physique des particules, optique corpusculaire, in Electron Beams,
Lenses, and Optics, Université d’Orsay 1966–1967, vol. I, ed. by J.C. El-Kareh (Academic,
New York and London, 1970), pp.38–39
23. A. Galejs, P.H. Rose, Optics of electrostatic tubes, in Focusing of Charged Particles, vol. 2, ed.
by A. Septier (Academic, 1967)
24. F. Méot, Spin dynamics, in Polarized Beam Dynamics and Instrumentation in Particle Accel-
erators, USPAS Summer 2021 Spin Class Lectures. (Springer Nature, Open Access, 2023).
https://link.springer.com/book/10.1007/978-3-031-16715-7
25. G.K. Green, E.E. Courant, The electron analog, in Handbuch der Physik, vol. XLIV, (Springer,
Berlin, 1959), p.319
References 53

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0
International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing,
adaptation, distribution and reproduction in any medium or format, as long as you give appropriate
credit to the original author(s) and the source, provide a link to the Creative Commons license and
indicate if changes were made.
The images or other third party material in this chapter are included in the chapter’s Creative
Commons license, unless indicated otherwise in a credit line to the material. If material is not
included in the chapter’s Creative Commons license and your intended use is not permitted by
statutory regulation or exceeds the permitted use, you will need to obtain permission directly from
the copyright holder.
Chapter 3
Classical Cyclotron

Abstract This chapter introduces the classical cyclotron, and the theoretical material
needed for the simulation exercises. It begins with a brief reminder of the historical
context, and continues with beam optics and with the principles and methods which
the classical cyclotron leans on, including
– ion orbit in a cyclic accelerator,
– weak focusing and periodic transverse motion,
– revolution period and isochronism,
– voltage gap and resonant acceleration,
– the cyclotron equation.
The simulation of a cyclotron dipole will either resort to an analytical model of the
field: the optical element DIPOLE, or will resort to using a field map together with
the keyword TOSCA to handle it and raytrace through. An additional accelerator
device needed in the exercises, CAVITE, simulates a local oscillating voltage. Run-
ning a simulation generates a variety of output files, including the execution listing
zgoubi.res, always, and other zgoubi.plt, zgoubi.CAVITE.out, zgoubi.MATRIX.out,
etc., aimed at looking up program execution, storing data for post-treatment, produc-
ing graphs, etc. Additional keywords are introduced as needed, such as the match-
ing procedure FIT[2]; FAISCEAU and FAISTORE which log local particle data in
zgoubi.res or in a user defined ancillary file; MARKER; the “system call” command
SYSTEM; REBELOTE, a ‘do loop’; and some more. This chapter introduces in addi-
tion to spin motion in accelerator magnets; dedicated simulation exercises include a
variety of keywords: SPNTRK, a request for spin tracking, SPNPRT or FAISTORE,
to log spin vector components in respectively zgoubi.res or some ancillary file, and
the “IL = 2” flag to log stepwise particle data, including spin vector, in zgoubi.plt
file. Simulations include deriving transport matrices, beam matrix, optical functions
and their transport, from rays, using MATRIX and TWISS keywords.

© The Author(s) 2024 55

F. Méot, Understanding the Physics of Particle Accelerators, Particle Acceleration
and Detection, https://doi.org/10.1007/978-3-031-59979-8_3
56 3 Classical Cyclotron

Notations Used in the Text

. B; . B0 magnetic field; at a reference radius . R0

.B; . B R ; . B y field vector; radial component; axial component
. B R = p/q magnetic rigidity
.C; .C0 orbit length, .C = 2π R; reference, .C0 = 2π R0
.E ion energy, . E = γ m 0 c2
. f rev , . f rf revolution and RF voltage frequencies
.G gyromagnetic anomaly, .G = 1.7928 for proton, .−4.184 for helion
.h harmonic number, an integer, .h = f rf / f rev
.k =
R dB
B dR
radial field index
.m; .m 0 ; . M ion mass; rest mass; in units of MeV/c.2
.p; . p; . p0 ion momentum vector; its modulus; reference
.q ion charge
. R; R0 ; R E equilibrium orbit radius; reference, . R( p0 ); at energy E
.R F Radio-Frequency
.s path variable
. Trev , . Trf revolution and accelerating voltage periods
.v; .v ion velocity vector; its modulus
. V (t); V̂ oscillating voltage; its peak value
.W kinetic energy, .W = 21 mv2
[ ]
.x ,, y, radial and axial coordinates . (∗), = d(∗)
ds
.α trajectory deviation, or momentum compaction
.β = ; .β0 ; .βs
v
c
normalized ion velocity; reference; synchronous
.γ = E/m 0 c
2
Lorentz relativistic factor
./\p, .δp momentum offset
.εu Courant-Snyder invariant (.u : x, r, y, l, Y, Z , s, etc.)
.θ azimuthal angle
.φ RF phase at ion arrival at the voltage gap

3.1 Introduction

Cyclotrons are the most widespread type of accelerator, today, used by thousands,
with the production of isotopes as the dominant application. This chapter is devoted
to the first cyclic accelerator: the early 1930s classical cyclotron which its concept
limited to low energy, a few tens of MeV/nucleon. This limitation was overcome a
decade later by the azimuthally varying field (AVF) technique, this is the subject of
the next chapter.
The classical cyclotron is based on four main principles:
(i) the use of a cylindrical-symmetry magnetic field in the gap of an electromagnet
(Fig. 3.1) to maintain ions on a circular trajectory
3.1 Introduction 57

MAGNET YOKE

d
acte h
extr bunc

separation
electrode

accelerating gap
V(t)

Fig. 3.1 Left: a cyclotron electromagnet, namely here that used for a model of Berkeley’s 184
inch cyclotron in the early 1940s [3]. Magnetic field in the gap decreases with radius. Right: a
schematic of the resonant acceleration motion; turn after turn, accelerated ions spiral out (bottom)
in the quasi-uniform field (top). A double-dee (or, a variant, a single-dee facing a slotted electrode)
forms an accelerating gap. The fixed-frequency oscillating voltage.V (t) applied is a harmonic of the
revolution frequency. Ions experiencing proper voltage phase at the gap, turn by turn, are accelerated.
A septum electrode allows beam extraction

(ii) transverse vertical confinement of the beam obtained by a slow radial decrease
of the magnetic field. A technique known as weak focusing, applied over the
years in all cyclic accelerators: microtron, betatron, synchrocyclotron, syn-
chrotron. These weak focusing accelerator species all are still part of the land-
scape today
(iii) resonant acceleration by synchronization of a fixed-frequency accelerating volt-
age on the quasi-constant revolution time (Fig. 3.1) and
(iv) use of high voltage, to mitigate the effect of the turn-by-turn RF phase slip.
Resonant acceleration has the advantage that a small gap voltage is enough to
accelerate with, in principle, no energy limitation, by contrast with the electrostatic
techniques developed at the time, which required the generation of the full voltage,
such as the Van de Graaff which was limited by sparking at a few tens of megavolts.
The cyclotron concept goes back to the late 1920s [1], yet it was not until the early
1930s when a cyclotron was first brought to operation [2]. The principles are sum-
marized in Fig. 3.1: an oscillating voltage is applied on a pair of electrodes (“dees”)
forming an accelerating gap and placed between the two poles of an electromagnet.
Ions reaching the gap during the acceleration phase of the voltage wave experience
an energy boost; no field is experienced inside the dees. Under the effect of energy
increase at the gap every half-revolution, they spiral out in the quasi-constant field
of the dipole.
The first cyclotron achieved acceleration of . H2+ hydrogen ions to 80 keV [2],
at Berkeley in 1931. The apparatus used a dee-shaped electrode vis-à-vis a slotted
electrode forming a voltage gap, the ensemble housed in a 5 in diameter vacuum
chamber and placed in the 1.3 Tesla field of an electromagnet. A .≈ 12 MHz vacuum
tube oscillator provided 1 kVolt gap voltage.
58 3 Classical Cyclotron

Fig. 3.2 Berkeley 27 inch cyclotron, brought to operation in 1934, accelerated deuterons up to
6 MeV. Left: a double-dee (seen in the vacuum chamber, cover off), 22 in diameter, creates an
accelerating gap: 13 kV, 12 MHz radio frequency voltage is applied for deuterons for instance
(through two feed lines seen at the top right corner). This apparatus was dipped in the 1.6 Tesla
dipole field of a 27 in diameter, 75 ton, electromagnet. A slight decrease of the dipole field with
radius, from the center of the dipole, ensures axial beam focusing. With their energy increasing,
ions spiral out from the center to eventually strike a target (red arrow). Right: ionization of the air
by the extracted beam (1936); the view also shows the vacuum chamber squeezed between the pole
pieces of the electromagnet [3]

Fig. 3.3 Berkeley 184 in

diameter, 4,000 ton cyclotron
during construction [3]. The
coil windings around both of
the magnetic poles are
clearly visible. Following the
invention of longitudinal
focusing it was actually
operated as a
synchrocyclotron, in 1946.
The man on the right gives
the scale

One goal foreseen in developing this technology was the acceleration of protons
to MeV energy range for the study of atom nucleus. And in background, a wealth
of potential applications. An 11 in cyclotron followed which delivered a 0.01.µA
+
. H2 beam at 1.22 MeV [4], and a 27 in cyclotron later reached 6 MeV (Fig. 3.2) [5].
Targets were mounted at the periphery of the 11 inch cyclotron, disintegrations were
observed in 1932. And, in 1933: ‘The neutron had been identified by Chadwick
in 1932. By 1933 we were producing and observing neutrons from every target
bombarded by deuterons.’ [5, M. S. Livingston, p. 22].
A broad range of applications were foreseen: “At this time biological experiments
were started. [...] Also at about this same time the first radioactive tracer experiments
on human beings were tried [...] simple beginnings of therapeutic use, coming a
3.1 Introduction 59

Fig. 3.4 Evolution of the

number of the various
cyclotron species, over the
years [9]. From the 1950s on
the AVF cyclotron rapidly
supplanted the 1930s’
classical cyclotron

little bit later, in which neutron radiation was used, for instance, in the treatment
of cancer. [...] Another highlight from 1936 was the first time that anyone tried
to make artificially a naturally occurring radio-nuclide. (a bismuth isotope) ” [5,
McMillan, p. 26].
Berkeley’s 184 in cyclotron, the largest (Fig. 3.3), commissioned in 1941, was
to accelerate Deuterons to 100 MeV for meson production. Its magnet however was
diverted to the production of uranium for the atomic bomb during the second world
war years [1]. Re-started in 1946, as a consequence of the discovery of phase focusing
the accelerator was actually operated as a synchrocyclotron (an accelerator species
addressed in Chap. 7).
Limitation in energy
The understanding of the dynamics of ions in the classical cyclotron took some time,
and brought two news, a bad one and a good one,
(i) the bad one first: the energy limitation. A consequence of the loss of isochronism
resulting from the relativistic increase of the ion mass so that “[...] it seems
useless to build cyclotrons of larger proportions than the existing ones [...] an
accelerating chamber of 37 in radius will suffice to produce deuterons of 11 MeV
energy which is the highest possible [...]” [6], or in a different form: “If you went
to graduate school in the 1940s, this inequality (.−1 < k < 0) was the end of the
discussion of accelerator theory ” [7].
(ii) the good news now: the energy limit which results from the mass increase can be
removed by shaping the magnetic pole into valley and hill field sectors. This is
the azimuthally varying field (AVF) cyclotron technology, due to L.H. Thomas
in 1938 [8]. It took some years to see effects of this breakthrough (Fig. 3.4). The
AVF is the object of Chap. 4.
60 3 Classical Cyclotron

With the progress in magnet computation tools, in computer speed and in beam
dynamics simulations, the AVF cyclotron ends up being essentially as simple to
design and build: it has in a general manner supplanted the classical cyclotron in all
energy domains (Fig. 3.4).

3.2 Basic Concepts and Formulæ

The cyclotron was conceived as a means to overcome the technological difficulty of

a long series of high electrostatic voltage electrodes in a linear layout, by, instead,
repeated recirculation through a single accelerating gap in synchronism with an
oscillating voltage (Fig. 3.5). As the accelerated bunch spirals out in the uniform
magnetic field, the velocity increase comes with an increase in orbit length; the
net result is a slow increase of the revolution period .Trev with energy, yet, with
appropriate fixed . f rf ≈ h/Trev the revolution motion and the oscillating voltage can
be maintained in sufficiently close synchronism, .Trev ≈ Trf / h, that the bunch will
transit the voltage gap at an accelerating phase (Fig. 3.6) over a large enough number
of turns that it acquires a significant energy boost.
++++++++

− − − − − − − −

A E
h=1
V>0
− − − − − − − −

++++++++++

h=3
A’
V<0

Fig. 3.5 Resonant acceleration: in an.h = 1 configuration an ion bunch meets an oscillating field.E
across gap A, at time .t, at an accelerating phase; it meets again, half a turn later, at time .t + Trev /2,
the accelerating phase across gap A’, and so on: the magnetic field recirculates the bunch through
the gap, repeatedly. Higher harmonic allows more bunches: the next possibility in the present
configuration is h = 3, and 3 bunches, 120.◦ apart, in synchronism with .E
3.2 Basic Concepts and Formulæ 61

V=V sin( ωRFt)

Fig. 3.6 An ion which
reaches the double-dee gap
at the RF phase .ωrf t = φ A or
.ωrf t = φ B is accelerated. If
it reaches the gap at
.ωrf t = φC it is decelerated

φC φ=ωRFt
φA φB

The orbital motion quantities: radius . R, ion rigidity . B R, revolution frequency

f , satisfy
. rev

p v qB qB
. BR = , 2π f rev = ωrev = = = (3.1)
q R m γ m0

These relationships hold at all .γ , so covering the classical cyclotron domain (.v << c,
γ ≈ 1) as well as the isochronous cyclotron (in which the ion energy increase is
.

commensurate with its mass). To give an idea of the revolution frequency, in the
limit .γ = 1, for protons, one has . f rev /B = q/2π m = 15.25 MHz/T.
The cyclotron design sets the constant RF frequency . f rf = ωrf /2π at an interme-
diate value of .h f rev along the acceleration cycle. The energy gain, or loss, by the ion
when transiting the gap, at time .t, is

./\W (t) = q V̂ sin φ(t) with φ(t) = ωrf t − ωrev t + φ0 (3.2)

with .φ its phase with respect to the RF signal at the gap (Fig. 3.6), .φ0 = φ(t = 0),
and .ωrev t the orbital angle. Assuming constant field . B, the increase of the revolution
period with ion energy satisfies

/\Trev
. =γ −1 (3.3)
Trev

The mis-match so induced between the RF and cyclotron frequencies is a turn-by-turn

cumulative effect and sets a limit to the tolerable isochronism defect, ./\Trev /Trev ≈
2–3%, or highest velocity .β = v/c ≈ 0.22. This results for instance in a practical
limitation to .≈25 MeV for protons, and .≈50 MeV for D and .α particles, a limit
however dependent on energy gain per turn.
Over time multiple-gap accelerating structures where developed, whereby a
“multiple-./\” electrode pattern substitutes to a “double-D”. An example is GANIL C0
injector with its 4 accelerating gaps and.h = 4 and.h = 8 RF harmonic operation [10].
62 3 Classical Cyclotron

3.2.1 Fixed-Energy Orbits, Revolution Period

In a laboratory frame (O; x, y, z), with (O; x, z) the bend plane (Fig. 3.7), assume
B| y=0 = B y , constant. An ion is launched from the origin with a velocity
.

( )
d x dy dz
.v= , , = (v sin α, 0, v cos α)
dt dt dt

at an angle .α from the .z-axis. Solving

m v̇ = qv × B
. (3.4)

with .B = (0, B y , 0) yields the parametric equations of motion

⎧ v v cos α
⎪
⎪ x(t) = cos(ωrev t − α) −
⎪
⎨ ωrev ωrev
. y(t) = constant (3.5)
⎪
⎪ v v sin α
⎪
⎩ z(t) = sin(ωrev t − α) +
ωrev ωrev

which result in
( )2 ( )2 ( )2
v cos α v sin α v
. x+ + z− = (3.6)
ωrev ωrev ωrev

α v sin α
a circular trajectory of radius . R = v/ωrev centered at .(xC , z C ) = .(− v ωcos
rev
, ωrev ).

Stability of the cyclic motion—The initial velocity vector defines a reference closed
orbit in the median plane of the cyclotron dipole; a small perturbation in .α or .v results
in a new orbit in the vicinity of the reference. An axial velocity component .v y on
the other hand, causes the ion to drift away from the reference, vertically, linearly

Fig. 3.7 Circular motion of

an ion in the plane normal to
a uniform magnetic field .B. y
The orbit is centered at
x z
. x C = −v cos α/ωrev , α
.z C = v sin α/ωrev , its radius V B
is .v/ωrev O
zc xc
C
3.2 Basic Concepts and Formulæ 63

with time, as there is no axial restoring force. The next Section will investigate the
necessary field property to ensure both horizontal and vertical confinement of the
cyclic motion in the vicinity of a reference orbit in the median plane.

3.2.2 Weak Focusing

In the early accelerated turns in a classical cyclotron (central region of the electro-
magnet, energy up to tens of keV/u), the accelerating electric field provides vertical
focusing for particles with proper RF phase [11, Sect. 8], whereas a flat magnetic
field with uniformity .d B/B < 10−4 is sufficient to maintain isochronism. Beyond
this low energy region however, at greater radii, a magnetic field gradient must be
introduced to ensure transverse stability: field must decrease with . R.
Ion coordinates in the following are defined in the moving frame .(M0 ; s, x, y)
(Fig. 3.8), which moves along the reference orbit (radius . R0 ), with its origin . M0 the
location of the reference ion on the orbit; the .s axis is tangent to the latter, the .x axis
is normal to .s, the . y axis is normal to the bend plane. Median-plane symmetry of the
field is assumed, thus the radial field component . B R | y=0 = 0 at all . R (Fig. 3.9).
Consider small motion excursions .x(t) = r (t) − R0 << R0 ; introduce Taylor
expansion of the field components,
| | |
∂ B y || x 2 ∂ 2 B y || ∂ B y ||
. B y (R0 + x) = B y (R0 ) + x + + · · · ≈ B y (R0 ) + x
∂ R | R0 2! ∂ R 2 | R0 ∂ R | R0
| | |
∂ B R || y 3 ∂ 3 B R || ∂ B y ||
B R (0 + y) = y | + | + ··· ≈ y (3.7)
∂ y 0 3! ∂ y 0 3 ∂ R | R0
, ,,| ,
∂ By |
= ∂ R |R
0

Fig. 3.8 Moving frame y

.(M0 ; s, x, y) along the
reference circular orbit. The refe
ren
M x
ce
curvature .1/R0 is constant
along the orbit and v
.(M0 ; s, x, y) can be
considered equivalent to the r(s) M0
cylindrical frame
.(C; θ, R0 , y) s
R0
B
C
64 3 Classical Cyclotron

Fig. 3.9 Axial motion

stability requires proper y th
shaping of field lines: . B y has t po le, Sou
to decrease with radius. The Magne
Laplace force pulls a positive I B B=B y

g(r)
charge with velocity pointing F
out of the page, at I, toward Median
plane
the median plane. Increasing
F r
the field gradient (k closer to I B
.−1, gap opening up faster)
Magne
increases the focusing t pole,
North

˙ = d(∗)/dt, the linear approximation of the differential

Using these, and noting .(∗)
equations of motion in the moving frame writes
( | ) ( )
mv2 ∂ B y || mv2 x
. Fx = m ẍ = −qv B y (R) + ≈ −qv B y (R0 ) + x + 1 −
R0 + x ∂ R | R0 R0 R0
( | )
mv2 R0 ∂ B y ||
→ m ẍ = − 2 +1 x (3.8)
R0 B0 ∂ R | R0
|
∂ B R || ∂ By
Fy = m ÿ =qv B R (y) = qv | y + higher order → m ÿ = qv y
∂ y y=0 ∂R

Note . B y (R0 ) = B0 and introduce

( )
R0 ∂ B y 2 R0 ∂ B y
.ω R = ωrev 1+ , ω2y = −ωrev
2 2
(3.9)
B0 ∂ R B0 ∂ R

substitute in Eqs. 3.8, this yields

. ẍ + ω2R x = 0 and ÿ + ω2y y = 0 (3.10)

A restoring force (linear terms in x and y, Eq. 3.10) arises from the radially varying
field, characterized by a field index
|
R0 ∂ B y ||
k= (3.11)
B0 ∂ R | R=R0 ,y=0
.

Radial stability: radially this force adds to the geometrical focusing (curvature term
“1” in .ω2R , Eq. 3.9, Fig. 3.10). In the weakly decreasing field . B(R) an ion with
momentum . p = mv moving in the vicinity of the . R0 -radius reference orbit experi-
v2
ences in the moving frame a resultant force . Ft = −qv B + m (Fig. 3.11) of which
r
3.2 Basic Concepts and Formulæ 65

k=0
k<0
k>0

R0+δR
R0-δR

R0
0 0.05 0.1 0.15 0.2

Fig. 3.10 Geometrical focusing: take k = 0; two circular trajectories which start from.r = R0 ± δ R
(solid lines, going counter-clockwise) undergo exactly one oscillation around the reference orbit
.r = R0 . A negative k (triangles), for axial focusing, decreases the radial convergence; a positive k
(square markers) increases the radial convergence and increases vertical divergence

2
the (outward) component. f c = m vr decreases with r at a higher rate than the decrease
of the Laplace (inward) component . f B = −qv B(r ). In other words, radial stability
requires . B R to increase with . R, . ∂∂BRR = B + R ∂∂ BR > 0, this holds in particular at . R0 ,
thus .1 + k > 0.
Axial stability requires a restoring force directed toward the median plane. Refer-
ring to Fig. 3.9, this means. Fy = −a × y (with.a a positive quantity) and thus. B R < 0,
at all .(r, y /= 0). This is achieved by designing a guiding field which decreases with
radius, . ∂∂ByR < 0. Referring to Eq. 3.11 this means .k < 0.
From these radial and axial constraints the condition of “weak focusing” for
transverse motion stability around the circular equilibrium orbit results, namely,

. −1<k <0 (3.12)

Note regarding the geometrical focusing: the focal distance associated with the cur-
2
vature of a magnet of arc length .L is obtained by integrating . ddsx2 + R12 x = 0 and
0
identifying with the focusing property ./\x , = −x/ f , namely,
{ {
d2x −x −xL R2
./\x , = ds ≈ 2 ds = , thus f = (3.13)
ds 2 R R2 L

Isochronism: the axial focusing constraint, . B decreasing with . R, contributes

breaking the isochronism (in addition to the effect of the mass increase) by virtue of
.ωrev ∝ B.
66 3 Classical Cyclotron

y x

O
s
R0
R
C qvB I mv /R
2 increases
i e r
B
decreases
BR<mv/q BR= BR>mv/q
force toward I mv/q force toward I

Fig. 3.11 Radial motion stability. Trajectory arcs at . p = mv are represented: case of .k = 0 (thin
black lines), of.−1 < k < 0 (thick blue lines), and of.k = −1 (dashed concentric circles)..k decreas-
ing towards .−1 reduces the geometrical focusing, increases axial focusing. The resultant of the
Laplace and centrifugal forces, . Ft = −qv B + mv 2 /r , is zero at I, motion is stable if . Ft is toward
I at .i, i.e. .qv Bi < mv 2 /Ri , and toward I as well at .e, i.e. .qv Be > mv 2 /Re

Paraxial Transverse Coordinates

Introduce the path variable .s as the independent variable in Eq. 3.10 and neglect the
transverse velocity components (.1 + Rx0 ≈ 1, y << 0) so that

. (3.14)

thus the equations of motion in the moving frame (Eq. 3.10) take the form

d2x 1+k d2 y k
. + x =0 and − 2y =0 (3.15)
ds 2 R02 ds 2 R0

Given .−1 < k < 0 the motion is that of a harmonic oscillator, in both planes, with
respective restoring constants .(1 + k)/R02 and .−k/R02 , both positive quantities. The
solution is a sinusoidal motion,
3.2 Basic Concepts and Formulæ 67
{ √ √
r (s) − R0 = x(s) = x0 cos 1+k
R
(s − s0 ) + x0, √R1+k
0
sin R1+k (s − s0 )
. √ 0 √ 0√
(3.16)
,
r , (s) = x , (s) = −x0 R0 sin R0 (s − s0 ) + x0 cos R0 (s − s0 )
1+k 1+k 1+k

{ √ √
−k
y(s) = y0 cos R
(s − s0 ) + y0, √R−k
0
sin R−k (s − s0 )
. √ 0 √ 0√
(3.17)
y , (s) = −y0 R0 sin R0 (s − s0 ) + y0 cos R−k
−k −k ,
0
(s − s0 )

Radial and axial wave numbers can be introduced,

ωR √ ωy √
ν =
. R = 1 + k and ν y = = −k (3.18)
ωrev ωrev

i.e., the number of sinusoidal oscillations of the paraxial motion about the reference
circular orbit over a turn, respectively radial and axial. Both are less than 1: there
is less than one sinusoidal oscillation in a revolution. In addition, as a result of the
revolution symmetry of the field,

ν 2 + ν y2 = 1
. R (3.19)

Off-Momentum Orbit
In a structure with revolution
{ symmetry, the equilibrium trajectory at momentum
{ { ( )
p0 R0 with B0 R0 = pq0 B = B0 + ∂∂ Bx 0 /\x + · · ·
. is at radius . p , where .
p = p0 + /\p R with B R = q R = R0 + /\x
On the other hand
[ ( ) ]
p ∂B p0 + /\p
. BR = ⇒ B0 + /\x + · · · (R0 + /\x) =
q ∂x 0 q
[( ∂ B ) ]
which, neglecting terms in.(/\x)2 , and given. B0 R0 = pq0 , leaves./\x ∂x 0
R0 +B0 =
/\p ( )
q
. With .k = RB00 ∂∂ Bx 0 this yields

/\p R0
./\x = D with D= the dispersion function (3.20)
p0 1+k

The dispersion . D is an s-independent quantity as a result of the revolution symmetry

of the field (k and R = p/qB are s-independent).
To the first order in the coordinates, the vertical coordinates y(s), y’(s) (Eq. 3.17)
are unchanged under the effect of a momentum offset, the horizontal trajectory angle
x’(s) (Eq. 3.16) is unchanged as well (the circular orbits are concentric, Fig. 3.12)
whereas .x(s) satisfies
|
∂ x || /\p
x(s, p + /\p) = x(s, p ) + /\p = x(s, po ) + D (3.21)
∂ p |s, p0
. 0 0
p0
68 3 Classical Cyclotron

p0
y
pole
Magnet R0 p
.

R0 R R
R
A B
Magnet p
ole

Fig. 3.12 The equilibrium radius at location . A is . R0 , momentum is . p0 , rigidity is . B0 R0 . The

equilibrium radius at . B is . R, momentum . p, rigidity . B R

Orbit and revolution period lengthening

A .δp momentum offset results in (Eq. 3.20)

δC δR δx δp 1 1
. = = =α with α = = 2 (3.22)
C R R p 1+k νR

with .α the momentum compaction, a positive quantity: orbit length increases with
momentum. Substituting . δβ
β
= γ12 δpp , the change in revolution period .Trev = C/βc
with momentum writes
( )
δTrev δC δβ 1 δp
. = − = α− 2 (3.23)
Trev C β γ p

Given that.−1 < k < 0 and.γ > 1, it results that.α − 1/γ 2 > 0: the revolution period
increases with energy, the increase in radius is faster than the velocity increase.

3.2.3 Quasi-Isochronous Resonant Acceleration

The energy .W of an accelerated ion (in the non-relativistic energy domain of the
classical cyclotron) satisfies the frequency dependence
( )
1 2 1 1 f rf 2
.W = mv = m (2π R f rev ) = m 2π R
2
(3.24)
2 2 2 h

Observe in passing: given the cyclotron size (radius . R), . f rf and .h set the limit for the
acceleration range. The revolution frequency decreases with energy and the condi-
tion of synchronism with the oscillating voltage, . f rf = h f rev , is only fulfilled at that
3.2 Basic Concepts and Formulæ 69

0.15
g
rtin
stapoint
ω rf = qB/m Δφ
<0
0.1
B [T] isochronism
0.5

ΔW [MeV]
0.05
synchronism
at one point

Δφ
Δφ=0

>0
(frf=frev)
ω rev> ω rf ωrev< ωrf 0

Δφ < 0 Δφ > 0
-0.05

-0.1
0 50 100 R [cm] 0.5 1 1.5 2 2.5 3 3.5 4
RF phase φ [rad]

Fig. 3.13 Left: a sketch of the synchronism condition at one point (h = 1 assumed). Right: the span
in phase of the energy gain ./\W = q V̂ sin φ (Eq. 3.2) over the acceleration cycle

particular radius where .ωrf = q B/m (Fig. 3.13-left). The out-phasing ./\φ of the RF
at ion arrival at the gap builds-up turn after turn, decreasing in a first stage (towards
lower voltages in Fig. 3.13-right) and then increasing back to .φ = π/2 and beyond
towards .π. Beyond .φ = π the RF voltage is decelerating.
With .ωrev constant between two gap passages, differentiating .φ(t) (Eq. 3.2)
yields .φ̇ = ωrf − ωrev . Between two gap passages on the other hand, ./\φ = φ̇/\T =
φ̇Trev /2 = φ̇ πvR , yielding a phase-shift of
( ) ( )
ωrf mωrf
. half-turn /\φ = π −1 =π −1 (3.25)
ωrev (R) q B(R)

The out-phasing is thus a gap-after-gap, cumulative effect. Due to this the clas-
sical cyclotron requires quick acceleration (small number of turns), which means
high voltage (tens to hundreds of kVolts). As expected, with .ωrf and B constant,
.φ presents a minimum (.φ̇ = 0) at .ωrf = ωrev = q B/m where exact isochronism
is reached (Fig. 3.13). The upper limit to .φ is set by the condition ./\W > 0:
acceleration.
The cyclotron equation determines the achievable energy range, depending on
the injection energy . E i , the RF phase at injection .φi , the RF frequency .ωrf and gap
voltage .V̂ . It writes [12]
[ ]
ωrf E + E i E − Ei
. cos φ = cos φi + π 1 − (3.26)
ωrev 2M q V̂

Equation 3.26 is represented in Fig. 3.14 for various values of the peak voltage
and phase at injection .φi . . M [eV/c2 ] and . E [eV] are respectively the rest mass and
relativistic energy, .q V̂ is expressed in electron-volts, the index .i denotes injection
parameters.
70 3 Classical Cyclotron

Fig. 3.14 A graph of the φi =π/4

cyclotron equation 1
(Eq. 3.26), for three different
accelerating voltages: 100,
200 and 400 kV/gap
0.5
(respectively square, circle
and triangle markers). The φi =π/2

cos(φ)
sole settings resulting in
.−1 < cos φ(E) < 1, .∀E, 0
100 kV
allow complete acceleration
200 kV
to top energy. .φi = π/4 at
injection for instance, does -0.5 400 kV
not (upper three curves).
φi =3π/4
.φi = 3π/4 works (lower
three curves), with as low as
-1
100 kV/gap
0 5 10 15 20
W [MeV]

3.2.4 Beam Extraction

From . R = p/q B and assuming . B(R) ≈ constant (this is legitimate as .k is normally

small), in the non-relativistic approximation (.W << M, .W = p 2 /2M) one gets

dR 1 dW
. = (3.27)
R 2 W
Integrating yields
W
. R 2 = Ri2 (3.28)
Wi

with . Ri , .Wi initial conditions. From Eqs. 3.27 and 3.28, assuming .Wi << W and con-
stant acceleration rate.dW such that.W = n dW after n turns, one gets the scaling laws

√ R 1 dR R
. R∝ n, dR ∝ ∝ ∝ dW, = (3.29)
W R dn 2n
The turn separation .d R is proportional to the energy gain per turn and inversely
proportional to the orbit radius.
The radial distance between successive turns decreases with energy, toward zero
(Fig. 3.15), eventually resulting in insufficient spacing for insertion of an extraction
septum.

Orbit modulation
Consider an ion bunch injected in the cyclotron with some .(x0 , x0, ) conditions in
the vicinity of the reference orbit, and assume slow acceleration. While accelerated
the bunch undergoes an oscillatory motion around the equilibrium orbit (Eq. 3.16).
3.2 Basic Concepts and Formulæ 71

Fig. 3.15 The radial 0.8

distance between successive
turns decreases with energy, 0.6
in inverse proportion to the
orbit radius. The red and
0.4
blue segments here figure the
accelerating gap
0.2

YLab [m]
0
0 0.2 0.4 0.6 0.8

-0.2

-0.4

-0.6

-0.8
-0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8
XLab [m]

Observed at the extraction septum this oscillation modulates the distance of the bunch
to the local equilibrium orbit, moving it outwards or inwards depending on the turn
number, which modulates the distance between the accelerated turns. This effect
can be resorted to, so to increase the separation between the final two turns and so
enhance the extraction efficiency [9].

3.2.5 Spin Dance

“Much of the physics of spin motion can be illustrated using the simplest model of a
storage ring consisting of uniform horizontal bending and no straight sections” [13].
By virtue of this statement, a preliminary introduction to spin motion in magnetic
fields is given in the present chapter. In support to this in addition, comes the fact
that cyclotrons happened to be the first circular machines to acelerate polarized
beams (first acceleration of polarized beams had happened earlier in the 1960s, using
electrostatic columns at voltage generators, when polarized proton and deuteron
sources began operating [14]).
The magnetic field .B of the cyclotron dipole exerts a torque on the spin angular
momentum .S of an ion, causing it to precess following the Thomas-BMT differential
equation [15]
dS q [ ]
. =S × (1 + G)B|| + (1 + Gγ )B⊥ (3.30)
dt m
, ,, ,
ωsp
72 3 Classical Cyclotron

Fig. 3.16 Spin and velocity α

vector precession in a
constant field, from .S to .S, y x

S
and .v to .v, respectively. In
the moving frame the spin v
precession along the arc R
x
.L = Rα is .Gγ α, in the y

(1+G γ)α
laboratory frame the spin x’
precesses by .(1 + Gγ )α

Gγα
S’
v’

where .t is the time; .ωsp the precession vector: a combination of .B|| and .B⊥ compo-
nents of .B respectively parallel and orthogonal to the ion velocity vector. .G is the
gyromagnetic anomaly,
G = 1.7928474 (proton), –0.178 (Li), –0.143 (deuteron), –4.184 (.3 He) ...
.S in this equation is in the ion rest frame, all other quantities are in the laboratory

frame.
In the case of an ion moving in the median plane of the dipole, .B|| = 0, thus
the precession axis is parallel to the magnetic field vector, .B y , so that .ωsp = mq (1 +
Gγ )B y . The spin precession angle over a trajectory arc .L is
{ {
1 (L) B ds
θ
. sp, Lab = ωsp ds = (1 + Gγ ) = (1 + Gγ )α (3.31)
v (L) BR

with .α the velocity vector precession (Fig. 3.16). The precession angle in the moving
frame (the latter rotates by an angle .α along .L) is

θ = Gγ α
. sp (3.32)

thus the number of .2π spin precessions per ion orbit around the cyclotron is .Gγ . By
analogy with the wave numbers (Eq. 3.18) this defines the “spin tune”

ν = Gγ
. sp (3.33)

3.3 Exercises

Note: some of the input data files for these simulations are available in zgoubi source-
forge repository at https://sourceforge.net/p/zgoubi/code/HEAD/tree/
branches/exemples/book/zgoubiMaterial/cyclotron_classical/
3.3 Exercises 73

3.1 Modeling a Cyclotron Dipole: Using a Field Map

Solution 3.1
In this exercise, ion trajectories are ray-traced, various optical properties addressed
in the foregoing are recovered, using a field map to simulate the cyclotron dipole.
Fabricating that field map is a preliminary step of the exercise.
The interest of using a field map is that it is an easy way to account for fancy magnet
geometries and fields, including field gradients and possible defects. A field map can
be generated using mathematical field models, or from magnet computation codes, or
from magnetic measurements. The first method is used, here. TOSCA[MOD.MOD1
= 22.1] keyword [16, cf. INDEX] is used to ray-trace through the map.

Working hypotheses: A 2-dimensional.m(R, θ ) polar meshing of the median plane

is considered (Fig. 3.17). It is defined in a .(O; X, Y ) frame and covers an angular
sector of a few tens of degrees. The mid-plane field map is the set of values . B Z (R, θ )
at the nodes of the mesh. During ray-tracing, TOSCA[MOD.MOD1 = 22.1] extrapo-
lates the field along 3D space .(R, θ, Z ) ion trajectories from the 2D polar map [16].
(a) Construct a .180◦ two-dimensional map of a median plane field . B Z (R, θ ), proper
to simulate the field in a cyclotron as sketched in Fig. 3.1. Use one of the following
two methods: either (i) write an independent program, or (ii) use zgoubi and its
analytical field model DIPOLE, together with the keyword OPTIONS[CONSTY =
ON] [16, cf. INDEX].
Besides: use a uniform mesh (Fig. 3.17) covering from Rmin = 1 to Rmax = 76 cm,
with radial increment ./\R = 0.5 cm, azimuthal increment ./\θ = 0.5 [cm]/
R0 with . R0 some reference radius (say, 50 cm, in view of subsequent exercises), and
constant axial field. B Z = 5 kG. The appropriate 6-column formatting of the field map
data for TOSCA[MOD.MOD1 = 22.1] to read is the following: . R cos θ, Z , R sin θ,
BY, B Z , B X with .θ varying first, . R varying second; Z is the vertical direction
(normal to the map mesh), . Z ≡ 0 in the present case. Note that proper functioning of

Fig. 3.17 Principle of a 2D Y

field map in polar
coordinates, covering a .180◦ ΔR
sector (over the right hand
side dee). The mesh nodes Δθ
θ
.m(R, θ) are distant ./\R
radially, ./\θ azimuthally.
The map is used twice to X
cover the .360◦ cyclotron
dipole as sketched here, O
while allowing insertion of R
m(R, θ)
an accelerating gap between
the two dees
74 3 Classical Cyclotron

TOSCA requires the field map to begin with the following line of numerical values:
Rmin [cm] ./\R [cm] ./\θ [deg] . Z [cm]
Produce a graph of the . B Z (R, θ ) field map content.
(b) Ray-trace a few concentric circular mid-plane trajectories centered on the center
of the dipole, ranging in .10 ≤ R ≤ 80 cm. Produce a graph of these concentric tra-
jectories in the .(O; X, Y ) laboratory frame.
Initial coordinates can be defined using OBJET, particle coordinates along trajec-
tories during the stepwise ray-tracing can be logged in zgoubi.plt by setting IL =
2 under TOSCA. In order to find the Larmor radius corresponding to a particular
momentum, the matching procedure FIT can be used. In order to repeat the latter for
a series of different momenta, REBELOTE[IOPT = 1] can be used.
Explain why it is possible to push the ray-tracing beyond the 76 cm radial extent of
the field map.
(c) Compute the orbit radius . R and the revolution period .Trev as a function of kinetic
energy .W or rigidity . B R. Produce a graph, including for comparison the theoretical
dependence of .Trev .
(d) Check the effect of the density of the mesh (the choice of ./\R and ./\θ values,
◦
i.e., the number of nodes . Nθ × N R = (1 + 180 /\θ
) × (1 + 80/\R
cm
)), on the accuracy of
the trajectory and time-of-flight computation.
(e) Check the effect of the integration step size on the accuracy of the trajec-
tory and time-of-flight computation, by considering a small ./\s = 1 cm and a large
./\s = 10 cm, at 200 keV and 5 MeV (proton), and comparing with theory.
(f) Consider a periodic orbit, thus its radius R should remain unchanged after step-
wise integration of the motion over a turn. However, the size ./\s of the numerical
integration step has an effect on the final value of the radius:
For two different cases, 200 keV (a small orbit) and 5 MeV (a larger one), provide a
graph of the dependence of the relative error .δ R/R after one turn, on the integration
step size ./\s (consider a series of ./\s values in a range ./\s : 0.1 mm → 20 cm).
REBELOTE[IOPT = 1] do-loop can be used to repeat the one-turn raytracing with
different ./\s.
3.2 Modeling a Cyclotron Dipole: Using an Analytical Field Model
Solution 3.2
This exercise is similar to Exercise 3.1, yet using the analytical modeling DIPOLE,
instead of a field map. DIPOLE provides the Z-parallel median plane field .B(R, θ,
Z = 0) ≡ B Z (R, θ, Z = 0) at the projected.m(R, θ, Z = 0) ion location (Fig. 3.18),
while .B(R, θ, Z ) at particle location is obtained by extrapolation.
(a) Simulate a .180◦ sector dipole; DIPOLE requires a reference radius
[16, Eqs. 6.3.19–6.3.21], noted . R0 here; for the sake of consistency with other exer-
cises, it is suggested to take . R0 = 50 cm. Take a constant axial field . B Z = 5 kG.
Explain the various data that define the field simulation in DIPOLE: geometry, role
of . R0 , field and field indices, fringe fields, integration step size, etc.
Produce a graph of . B Z (R, θ ).
(b) Repeat question (b) of Exercise 3.1.
(c) Repeat question (c) of Exercise 3.1.
3.3 Exercises 75

Fig. 3.18 DIPOLE provides

the value . B Z (m) of the z M
median plane field at m,
projection of particle
position . M(R, θ, Z ) in the
median plane. .B(R, θ, Z ) is
obtained by extrapolation
BZ

R m
θ
θ= 0

(d) As in question (e) of Exercise 3.1, check the effect of the integration step size on
the accuracy of the trajectory and time-of-flight computation.
Repeat question (f) of Exercise 3.1.
(e) From the two series of results (Exercise 3.1 and the present one), comment on
various pros and cons of the two methods, field map versus analytical field model.
3.3 Resonant Acceleration
Solution 3.3
Based on the earlier exercises, using indifferently a field map (TOSCA) or an ana-
lytical model of the field (DIPOLE), introduce a sinusoidal voltage between the two
dees, with peak value 100 kV. Assume that ion motion does not depend on RF phase:
the boost through the gap is the same at all passes, use CAVITE[IOPT = 3] [16,
cf. INDEX] for that. Note that using CAVITE requires prior PARTICUL in order to
specify ion species and data, necessary to compute the energy boost (Eq. 3.2).
(a) Accelerate a proton with initial kinetic energy .20 keV, up to 5 MeV, take harmonic
h = 1. Produce a graph of the accelerated trajectory in the laboratory frame.
(b) Provide a graph of the proton momentum . p and total energy . E as a function of its
kinetic energy, both from this numerical experiment (ray-tracing data can be stored
using FAISTORE) and from theory, all on the same graph.
(c) Provide a graph of the normalized velocity.β = v/c as a function of kinetic energy,
both numerical and theoretical, and in the latter case both classical and relativistic.
(d) Provide a graph of the relative change in velocity ./\β/β and orbit length ./\C/C
as a function of kinetic energy, both numerical and theoretical. From their evolution,
conclude that the time of flight increases with energy.
(e) Repeat the previous questions, assuming a harmonic h=3 RF frequency.
3.4 Spin Dance
Solution 3.4
Cyclotron modeling in the present exercise can use Exercises 3.1 or 3.2 technique
(i.e., a field map or an analytical field model), indifferently.
76 3 Classical Cyclotron

(a) Add spin transport, using SPNTRK [16, cf. INDEX]. Produce a listing
(zgoubi.res) of a simulation, including spin outcomes.
Note: PARTICUL is necessary here, for the spin equation of motion (Eq. 3.30) to
be solved [16, Sect. 2]. SPNPRT can be used to have local spin coordinates listed in
zgoubi.res (at the manner that FAISCEAU lists local particle coordinates).
(b) Consider proton case, take initial spin longitudinal, compute the spin precession
over one revolution, as a function of energy over a range 12 keV.→5 MeV. Give a
graphical comparison with theory.
FAISTORE can be used to store local particle data, which include spin coordinates,
in a zgoubi.fai style output file. IL = 2 [16, cf. INDEX] (under DIPOLE or TOSCA,
whichever modeling is used) can be used to obtain a print out of particle and spin
motion data to zgoubi.plt during stepwise integration.
(c) Inject a proton with longitudinal initial spin .Si . Give a graphic of the longitudinal
spin component value as a function of azimuthal angle, over a few turns around the
ring. Deduce the spin tune from this computation. Repeat for a couple of different
energies.
Place both FAISCEAU and SPNPRT commands right after the first dipole sector,
and use them to check the spin rotation and its relationship to particle rotation, right
after the first passage through that first sector.
(d) Spin dance: the input data file optical sequence here is assumed to model a full
turn. Inject an initial spin at an angle from the horizontal plane (this is in order to
have a non-zero vertical component), produce a 3-D animation of the spin dance
around the ring, over a few turns.
(e) Repeat questions (b–d) for two additional ions: deuteron (much slower spin pre-
cession), .3 H e2+ (much faster spin precession).

3.5 Synchronized Spin Torque

Solution 3.5
A synchronized spin kick is superimposed on orbital motion. An input data file for a
complete cyclotron is considered as in question Exercise 3.4(d), for instance six 60.◦
DIPOLEs, or two 180.◦ DIPOLEs.
Insert a local spin rotation of a few degrees around the longitudinal axis, at the
end of the optical sequence (i.e., after one orbit around the cyclotron). SPINR can be
used for that, rather than a local magnetic field, so to avoid any orbital effect. Track
4 particles on their respective equilibrium orbit, with energies 0.2, 108.412, 118.878
and 160.746 MeV.
Produce a graph of the motion of the vertical spin component . S y along the circular
orbit.
Produce a graph of the spin vector motion on a sphere.

3.6 Weak Focusing

Solution 3.6
(a) Consider a 60.◦ sector as in earlier exercises (building a field map and using
TOSCA as in Exercise 3.1, or using DIPOLE as in Exercise 3.2), construct the sector
3.3 Exercises 77

accounting for a non-zero radial index .k in order to introduce axial focusing, say
.k = −0.03, assume a reference radius . R0 for a reference energy of 200 keV (. R0 and
. B0 are required in order to define the index k, Eq. 3.11). Ray-trace that 200 keV
reference orbit, plot it in the lab frame: make sure it comes out as expected, namely,
constant radius, final and initial angles zero.
(b) Using FIT[2], find and plot the radius dependence of orbit rigidity, . B R(R), from
ray-tracing over a . B R range covering 20 keV to 5 MeV; superpose the theoretical
curve. REBELOTE[IOPT = 1] can be used to perform the scan.
(c) Produce a graph of the paraxial axial motion of a 1 MeV proton, over a few turns
(use IL = 2 under TOSCA, or DIPOLE, to have step by step particle and field data
logged in zgoubi.plt). Check the effect of the focusing strength by comparing the
trajectories for a few different index values, including close to –1 and close to 0.
(d) Produce a graph of the magnetic field experienced by the ion along these trajec-
tories.

3.7 Loss of Isochronism

Solution 3.7
Compare on a common graphic the revolution period .Trev (R) for a field index value
k ≈ −0.95, −0.5, −0.03, 0. The scan method of Exercise 3.6, based on REBE-
.
LOTE[IOPT=1] preceded by FIT[2], can be referred to.

3.8 Ion Trajectories

Solution 3.8
In this exercise individual ion trajectories are computed. DIPOLE or TOSCA mag-
netic field modeling can be used, indifferently. Take for instance . B0 = 5 kG and
for reference . R0 the 200 keV radius. No acceleration here, ions circle around the
cyclotron at constant energy.
(a) Produce a graph of the horizontal .x(s) and vertical . y(s) trajectory coordinates
of an ion with rigidity . B R(R0 ) and paraxial motion, over a few turns around the
cyclotron. From the number of turns, give an estimate of the wave numbers. Check
the agreement with the expected .ν R (k), .ν y (k) values (Eq. 3.18).
(b) Consider now protons in that very cyclotron, far from the reference energy. E(R0 ),
say at 1 MeV and 5 MeV. The wave numbers change with energy: compute k(E) from
tracking and check consistency with theory.
(c) Consider a proton, 200 keV energy, plot as a function of .s the difference between
. x(s) from raytracing and its values from Eq. 3.16. Same for y(s) compared to Eq. 3.17.
IL = 2 can be used to store in zgoubi.plt the step-by-step particle coordinates across
DIPOLE.
(d) Perform a scan of the wave numbers over 200 keV.−5 MeV energy interval,
computed using OBJET[KOBJ = 5] and MATRIX[IORD = 1, IFOC = 11], or
OBJET[KOBJ = 6] and MATRIX[IORD = 2, IFOC = 11], together with REBE-
LOTE[IOPT = 1] to repeat MATRIX for a series of energy values.
78 3 Classical Cyclotron

3.9 RF Phase at the Accelerating Gap

Solution 3.9
Consider the cyclotron model of exercise 3.6: field index .k = −0.03 defined at . R0 =
50 cm, field . B0 = 5 kG on that radius, two dees, double accelerating gap.
Accelerate a proton from 1 to 5 MeV: get the turn-by-turn phase-shift at the gaps;
use CAVITE[IOPT=7] to simulate the acceleration. Compare the .half-turn /\φ so
obtained with the theoretical expectation (Eq. 3.25). Produce similar graphs . B(R)
and ./\W (φ) to Fig. 3.13.
Accelerate over more turns, observe the particle decelerating.

3.10 The Cyclotron Equation

Solution 3.10
The cyclotron model of Exercise 3.3 is considered: two dees, double accelerating
gap, uniform field . B = 5 kG, no field gradient needed here (no vertical motion).

(a) Set up an input data file for the simulation of a proton acceleration from 0.2
to 20 MeV. In particular, assume that .cos(φ) reaches its maximum value at .Wm =
10 MeV; find the RF voltage frequency from .d(cos φ)/dW = 0 at .Wm .
(b) Give a graph of the energy-phase relationship (Eq. 3.26), for .φi = 3π
4
, π2 , π4 ,
from both simulation and theory.

3.11 Cyclotron Extraction

Solution 3.11
(a) Acceleration of a proton in a uniform field . B = 5 kG is first considered (field
hypotheses as in Exercise 3.3). RF phase is ignored: CAVITE[IOPT = 3] can be used
for acceleration. Take a 100 kV gap voltage.
Compute the distance ./\R between turns, as a function of turn number and of energy,
over the range . E : 0.02 → 5 MeV. Compare graphically with theoretical expecta-
tion.
(b) Assume a beam with Gaussian momentum distribution and rms momentum
spread .δp/ p = 10−3 . An extraction septum is placed half-way between two suc-
cessive turns, provide a graph of the percentage of beam loss at extraction, as a
function of extraction turn number. COLLIMA can be used for that simulation and
for particle counts, it also allows for possible septum thickness.
(c) Repeat (a) and (b) considering a field with index: take for instance . B0 = 5 kG
and .k = −0.03 at . R0 = R(0.2 MeV) = 12.924888 cm.
(d) Investigate the effect of injection conditions .(Yi , Ti ) on the modulation of the
distance between turns.
Try and confirm numerically that, with slow acceleration, the oscillation is minimized
x0 ν R
for an initial .|Ti | = | | (after Ref. [9, p. 133]).
R
3.4 Solutions of Exercises of This Chapter 79

3.12 Acceleration and Extraction of a 6-D Polarized Bunch

Solution 3.12
The cyclotron simulation hypotheses of Exercise 3.10a are considered; account or
k = −0.02 field index.
.

Add a short “high energy” extraction line, say 1 m, following REBELOTE in the
optical sequence, ending up with a “Beam_Dump” MARKER for instance.
(a) Create a 1,000 ion bunch with the following initial parameters:

– random Gaussian transverse phase space densities, centered on the equilibrium

orbit, truncated at 3 sigma, normalized rms emittances .εY = ε Z = 1 π μm, both
emittances matched to the 0.2 MeV orbit optics,
– uniform bunch momentum density .0.2 × (1 − 10−3 ) ≤ p ≤ 0.2 × (1 + 10−3 )
MeV, matched to the dispersion, namely (Eq. 3.21), ./\x = D /\p
p
,
– random uniform longitudinal distribution .−0.5 ≤ s ≤ 0.5 mm,

Note: two ways to create this object are, (i) using MCOBJET[KOBJ = 3] which
generates a random distribution, or (ii) using OBJET[KOBJ = 3] to read an exter-
nal particle coordinate file.

Add spin tracking request (SPNTRK), all initial spins normal to the bend plane.
Produce a graph of the three initial 2-D phase spaces: (Y, T), (Z, P), (.δl, .δp/ p),
matched to the 200 keV periodic optics.
Provide Y, Z, dp/p, .δl and . S Z histograms (HISTO can be used), check the distri-
bution parameters.

(b) Accelerate this polarized bunch to 20 MeV, using the following RF conditions:
– 200 kV peak voltage,
– RF harmonic 1,
– initial RF phase .φi = 3π/4.
Produce a graph of the three phase spaces as observed downstream of the extrac-
tion line. Provide the Y, Z, dp/p, .δl and . S Z histograms. Compare the distribution
parameters with the initial values.
What causes the spins to spread away from vertical?

3.4 Solutions of Exercises of This Chapter

3.1 Modeling a Cyclotron Dipole: Using a Field Map

(a) A field map of a .180◦ sector of a classical cyclotron magnet.
The first option is retained here: a Fortran program, geneSectorMap.f, given in
Table 3.1. constructs the required map of a field distribution . B Z (R, θ ), to be subse-
quently read and raytraced through using the keyword TOSCA [16, lookup INDEX].
80 3 Classical Cyclotron

Table 3.1 A Fortran program which generates a .180◦ mid-plane field map. This angle as well as
field amplitude can be changed, a field index can be added. This program can be compiled and run,
as is. The field map it produces is logged in geneSectorMap.out

Regarding the second option: using the analytical model DIPOLE together with
the keyword OPTIONS[CONSTY = ON] to fabricate a field map, examples can be
found for instance in the FFAG chapter exercises (Chap. 10).

A polar mesh is retained (Fig. 3.19), rather than Cartesian, consistently with
cyclotron magnet symmetry. The program can be compiled (gfortran-o geneSec-
torMap geneSectorMap.f will provide the executable, geneSectorMap) and run, as is.
The field map is saved under the name geneSectorMap.out, excerpts of the expected
content are given in Table 3.2. That name appears under TOSCA in zgoubi input
data file for this simulation (Table 3.3). Figure 3.20 shows the field over the .180◦
azimuthal extent (using a gnuplot script, bottom of Table 3.2).
Note the following:
(i) the field map azimuthal extent (set at .180◦ in geneSectorMap) can be changed,
for instance to simulate a 60 deg sector instead;
(ii) the field is vertical being the mid-plane field of dipole magnet. The field is taken
constant in this exercise, .∀R, ∀θ throughout the map mesh, whereas in upcom-
ing exercises, a focusing index will be introduced, which will make . B Z ≡ B Z (R)
an R-dependent quantity (in Chap. 4 which addresses Thomas focusing and the
isochronous cyclotron, exercises will further resort to . B Z ≡ B Z (R, θ ), an R- and
.θ -dependent quantity).
3.4 Solutions of Exercises of This Chapter 81

Fig. 3.19 Principle 2-D field map mesh as used by TOSCA, and the (O; X, Y) coordinate system.
A, B: Cartesian mesh in the (X, Y) plane, case of respectively 9-point and a 25-point interpolation
grid; the mesh increments are ./\X and ./\Y . C: polar mesh and increments ./\R and ./\α (./\θ in the
text), and (O; X, Y) frame moving along a reference arc of radius . R M . The field at particle location
is interpolated from its values at the closest .3 × 3 or .5 × 5 nodes

This field map can be readily tested using the example of Table 3.3, which raytraces
E k = 0.12, 0.2 and .5.52 MeV protons on circular trajectories centered at the center
.

of the field map. Trajectory radii, respectively . R = 10.011, 12.924 and .67.998 cm
(Table 3.3), have been prior determined from
/
.Rigidity Bρ = B0 × R and Bρ = p/c = E k (E k + 2 M)/c (3.34)

with . B0 = 0.5 T (Table 3.1) and . M = 938.272 MeV/c2 the proton mass.
The optical sequence for this particle raytracing uses the following keywords:

(i) OBJET to define a (arbitrary) reference rigidity and initial particle coordinates
(ii) TOSCA, to read the field map and raytrace through (and TOSCA’s ‘IL = 2’
flag to store step-by-step particle data into zgoubi.plt)
(iii) FAISCEAU to print out particle coordinates in zgoubi.res execution listing
(iv) SYSTEM to run a gnuplot script (Table 3.3) once raytracing is complete
82 3 Classical Cyclotron

Table 3.2 First and last few lines of the field map file geneSectorMap.out. The file starts with
an 8-line header, the first of which is effectively used by zgoubi (the following 7 are not used)
and indicates, in that order: the minimum radius of the map mesh Rmi, the radial increment dR,
the azimuthal increment dA, the axial increment dZ (null and not used in the present case of a
two-dimensional field map), in units of, respectively, cm, cm, degree, cm. The additional 7 lines
provide the user with various indications regarding numerical values used in, or resulting from, the
execution of geneSectorMap.f. The first 5 numerical data in line 5 in particular are to be reported
in zgoubi input data file under TOSCA keyword. The rest of the file is comprised of 8 columns,
the first three give the node coordinates and the next three the field component values at that node,
the last two columns are the (azimuthal and radial) node numbers, from (1, 1) to (315, 151) in the
present case

(v) MARKER, to define two particular “LABEL_1” type labels

[16, lookup INDEX] (#S_halfDipole and #E_halfDipole), to be used with INCLUDE
in subsequent exercises.
Three circular trajectories in a dee, resulting from the data file of Table 3.3 are
shown in Fig. 3.20. Inspecting zgoubi.res execution listing one finds the D, Y, T, Z,
P, S particle coordinates under FAISCEAU, at OBJET (left) and current (right) after
a turn in the cyclotron (unchanged, as the trajectory forms a closed orbit):

(b) Concentric trajectories in the median plane.

The optical sequence for this exercise is given in Table 3.4. Compared to the
previous sequence (Table 3.3), (i) the TOSCA segment has been replaced by an
INCLUDE, for the mere interest of making the input data file for this simulation
shorter, and (ii) additional keywords are introduced, including
3.4 Solutions of Exercises of This Chapter 83

Table 3.3 Simulation input data file FieldMapSector.inc: it is set to allow a preliminary test
regarding the field map geneSectorMap.out (as produced by the Fortran program geneSectorMap,
Table 3.1), by computing three circular trajectories centered on the center of the map. This file also
defines the INCLUDE segment between the labels (LABEL1 type [16, Sect. 7.7]) #S_halfDipole
and #E_halfDipole

0.8
"zgoubi.plt" u ($19==orbit ? $22 :1/0):($10 *cm2m):($19)
"zgoubi.plt" u ($19==orbit ? $22 :1/0):($10 *cm2m):($19)
"zgoubi.plt" u ($19==orbit ? $22 :1/0):($10 *cm2m):($19)
0.6

0.4

5.14
0.2
5.12
5.1
YLab [m]
B [kG]

5.08
5.06 0
0 0.2 0.4 0.6 0.8
5.04
5.02
5 -0.2
4.98

0.8 -0.4
0.7
0.6
0.5
-0.8 0.4
-0.6 -0.6
-0.4 0.3 X [m]
-0.2
0 0.2
0.2
Y [m] 0.4 0.1
0.6 -0.8
0.8 0
-0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8
XLab [m]

Fig. 3.20 Left: map of a constant magnetic field over a 180 deg sector, 76 cm radial extent. Right:
three circular trajectories, at respectively 0.12, 0.2 and 5.52 MeV, computed using that field map
84 3 Classical Cyclotron

Table 3.4 Simulation input data file: optical sequence to find cyclotron closed orbits at a series of
different momenta. An INCLUDE inserts the #S_halfDipole to #E_halfDipole TOSCA segment of
the sequence of Table 3.3

– FIT, which finds the circular orbit for a particular momentum,

– FAISCEAU, a means to check local particle coordinates,
– REBELOTE, which repeats the execution of the sequence (REBELOTE sends
the execution pointer back to the top of the data file) for a new momentum value
which REBELOTE itself defines, prior.
In order to compute and then plot trajectories (Fig. 3.21), zgoubi proceeds as
follows: orbit circles for a series of different radii taken in .[10, 80] cm are searched,
using FIT to find the appropriate momenta. REBELOTE is used to repeat that fitting
on a series of different values of R; prior to repeating, REBELOTE modifies the
initial particle coordinate .Y0 in OBJET. Stepwise particle data through the dipole
field are logged in zgoubi.plt, due to IL = 2 under TOSCA keyword, at the first pass
before FIT, and at the last pass following FIT completion. A key point here: a flag,
FITLST, recorded in column 51 in zgoubi.plt ([16], Sect. 8.3), is set to 1 at the last
pass (the last pass follows the completion of the FIT execution and uses updated FIT
variable values).
3.4 Solutions of Exercises of This Chapter 85

Fig. 3.21 Circular 1

"zgoubi.plt" u ($42==4*l && $51==1 ? $22 +pi*(l-1):1/0):($10 *cm2m):($49)
trajectories in the cyclotron "zgoubi.plt" u ($42==4*l && $51==1 ? $22 +pi*(l-1):1/0):($10 *cm2m):($49)
0.8
mid-plane, centered on the
field map center. The 0.6
outermost orbit is at R =
80 cm by hypothesis, thus 0.4
. B R = B0 × R = 0.4 T m,
0.2
. E k = 7.632 MeV. These

YLab [m]
stepwise .(R, θ) data are read 0
from zgoubi.plt, coordinates 0 0.2 0.4 0.6 0.8 1

(Y, X) in zgoubi polar frame -0.2

nomenclature (Sect. 8.3 [16])
-0.4

-0.6

-0.8

-1
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
XLab [m]

At the bottom of zgoubi input data file, a SYSTEM command produces a graph
of ion trajectories, by executing a gnuplot script (bottom of Table 3.4). Note the test
on FITLST, which allows selecting the last pass following FIT completion. Graphic
outcomes are given in Fig. 3.21.
The reason why it is possible to push the raytracing beyond the 76 cm radius field
map extent, without loss of accuracy, is that the field is constant. Thus, referring to
the polynomial interpolation technique used [16, Sect. 1.4], the extrapolation out of
the map will leave the field value unchanged.

(c) Energy and rigidity dependence of orbit radius and time-of-flight.

The orbit radius . R and the revolution time .Trev as a function of kinetic energy . E k
and rigidity . B R are obtained by a similar scan to exercise (b). The results are shown
in Fig. 3.22.
A slow increase of revolution period with energy can be observed, which is due
to the mass increase.
Note that these results are converged for the step size, to high accuracy (see
(d)), due to its value taken small enough, namely ./\s = 1 cm. This corresponds for
instance to 80 steps to complete a revolution for the 120 keV, . R = 12.9 cm smaller
radius trajectory in Fig 3.21.

(d) Numerical convergence: mesh density.

This question concerns the dependence of the numerical convergence of the solu-
tion of the differential equation of motion [16, Eq. 1.2.1] upon mesh density.
The program used in (b) to generate a field map (Table 3.1) is modified to construct
field maps of . B Z (R, θ ) with various radial and azimuthal mesh densities. Changing
these is simply a matter of modifying the quantities dR (radius increment ./\R) and
86 3 Classical Cyclotron

kin. E /MeV
0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
70 0.1319

0.1318
60

0.1317
50

0.1316

Trev /μs
R /cm

40

0.1315

30
0.1314

20
0.1313
R(B ρ) R(E kin ) TOF(B ρ)
theor. theor. theor.
10 0.1312
0.1 0.15 0.2 0.25 0.3
Bρ /T.m

Fig. 3.22 Numerical (markers) and theoretical (solid lines) values of orbit radius, R, and revolution
period, .Trev , versus kinetic energy (top scale) and rigidity (bottom scale). The mesh density here is
. Nθ × N R = 315 × 151. The integration step size is ./\s = 1 cm, so ensuring converged results (to
./\R/R and ./\Trev /Trev < 10 )
−6

R d A (R times the azimuth increment ./\θ) in the program of Table 3.1. The field
.
maps geneSectorMap.out so generated for various .(d R, Rd A) couples may be saved
under different names, and used separately.
Table 3.5 shows the complete, 9 line, TOSCA field map, in the case of a .60◦
60◦ 75 cm 360◦ 75 cm
sector covered in . Nθ × N R = × = ◦
× = 3 × 3 nodes. Six
/\θ /\R 120 37.5cm
sectors are now required to cover the complete cyclotron dipole: zgoubi input data
need be changed accordingly, namely stating TOSCA—possibly via an INCLUDE—
six times, instead of just twice in the case of a 180.◦ sector.
The result to be expected: with a mesh reduced to as low as . Nθ × N R = 3 ×
3, compared to . Nθ × N R = 106 × 151, radius and time-of-flight should however
remain unchanged. This shows in Fig. 3.23 which displays both cases, over a . E k :
0.12 → 5 MeV energy span (assuming protons). The reason for the absence of effect
of the mesh density is that the field is constant. As a consequence the field derivatives
in the Taylor series based numerical integrator are all zero [16, Sect. 1.2]: only . B Z
is left in evaluating the Taylor series, however . B Z is constant. Thus . R remains
unchanged when pushing the ion by a step ./\s, and the cumulated path length—
the closed orbit length—and revolution time—path length over velocity—end up
unchanged. Note: this will no longer be the case when a radial field index is introduced
in order to cause vertical focusing, in subsequent exercises.
3.4 Solutions of Exercises of This Chapter 87

Table 3.5 Field map of a .60◦ constant field sector as read by TOSCA. The field map is complete,
with smallest possible. Nθ × N R = 3 × 3 = 9 number of nodes. The first line of the header is used by
zgoubi (the following 7 are not used), namely, the minimum value of the radius in the map, radius
increment, azimuthal increment, and vertical increment (null here, as this is a single, mid-plane
map)

kin. E /MeV
0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
70 0.132

60 0.1315

50 0.131
Trev /μs

R(B ρ), Δs=10cm

R /cm

40 0.1305
3x3 mesh, Δs=1cm
theor.

30 R(E kin ), Δs=10cm 0.13

3x3 mesh, Δs=1cm
theor.
20 TOF(B ρ), Δs=10cm 0.1295
3x3 mesh, Δs=1cm
theor.
10 0.129
0.1 0.15 0.2 0.25 0.3
Bρ /T.m

Fig. 3.23 Convergence versus mesh density and step size: a graph of orbit radius . R (left axis), and
revolution period,.Trev (right axis), versus kinetic energy (top scale) and rigidity (bottom scale). Solid
markers are for ./\s = 1 cm and . Nθ × N R = 3 × 3 node mesh, large empty circles are for ./\s =
10 cm and . Nθ × N R = 106 × 151 node mesh. Solid lines are from theory and show convergence
in the case .3 × 3 nodes and ./\s = 1 cm
88 3 Classical Cyclotron

(e) Numerical convergence: integration step size

Figure 3.23 displays two cases of step sizes, ./\s ≈ 1 cm and ./\s = 10 cm.
It has been shown (Fig. 3.22) that ./\s ≈ 1 cm is small enough that the numerical
integration is converged, agreement with theoretical expectation is quite good.
The difference on the value of . R, in the case ./\s ≈ 10 cm, appears to be weak,
only noticeable at the scale of the graph for . R values small enough that the number of
steps over one revolution goes as low as .2π R//\s ≈ 2π × 14.5/10 ≈ 9. The change
in time-of-flight due to the larger step size amounts to a relative .10−3 .
Step size is critical in the numerical integration, the reason is that the coefficients
of the Taylor series that yield the new position vector .R(M1 ) and velocity vector
.v(M1 ), from an initial location . M0 after a ./\s push, are the derivatives of the velocity
vector [16, Sect. 1.2] and may take substantial values if .v(s) changes quickly. In
such case, taking too large a ./\s value makes the high order terms significant and
the Taylor series truncation [16, Eq. 1.2.4] is fatal to the accuracy (regardless of a
possible additional issue of radius of convergence of the series).

δR
(f) Numerical convergence: . (/\s)
R
Issues faced are the following:
– the increase of .δ R(/\s)/R at large ./\s has been addressed above;
– a small./\s is liable to cause an increase of.δ R(/\s)/R, due to computer accuracy:
truncation of numerical values at a limited number of digits may cause a ./\s push to
result in no change in .R(M1 ) (position) and .u(M1 ) (normed velocity) quantities [16,
Eq. 1.2.4].
A detailed answer to the question, including graphs, is left to the reader, the
method is the same as in (e).

3.2 Modeling a Cyclotron Dipole: Using an Analytical Field Model

This exercise introduces the analytical modeling of a dipole, using DIPOLE [16,
lookup INDEX], and compares outcomes to the field map case of Exercise 3.1. The
exercise is not entirely solved, however all the material needed for that is provided,
and indications are given to complete it.

(a) Analytical modeling.

DIPOLE keyword provides an analytical model of the field to simulate a sector
dipole with index, namely [16, lookup INDEX]
[ ( ) ( )2 ( )3 ]
R − R0 , R − R0 ,, R − R0
. B Z = F (θ )B0 1 + k +k +k (3.35)
R0 R0 R0

. R0 is a reference radius, . B0 = B Z (R0 )|F ≡1 is a reference field value, .k is the field

index and .k , , k ,, are homogeneous to its first and second derivative with respect to
3.4 Solutions of Exercises of This Chapter 89

R (Eq. 3.11). .F (θ ) is an azimuthal form factor, defined by the fringe field model,
presumably taking the value 1 in the body of the dipole. In the present case a hard-
edge field model is considered, so that
{
1 inside
F =
. the dipole magnet (3.36)
0 outside

Setting up the input data list under DIPOLE (Table 3.6) requires close inspection of
Fig. 3.24, which details the geometrical parameters such as the full angular opening
of the field region that DIPOLE comprises, AT; a reference angle ACN to allow
positioning the effective field boundaries at .ω+ and .ω− ; field and indices; fringe
field regions at . AC N − ω+ (entrance) and . AC N − ω− (exit); wedge angles, etc.
A 60 deg sector is used here for convenience, it is detailed in Table 3.6 (Table 3.7
provides the definition of a 180 deg sector, for possible comparisons with the present
three-sector assembly).
In setting up DIPOLE data the following values have been accounted for:
– . R0 = 50 cm, an arbitrary value (consistent with other exercises), more or less
half the dipole extent,
– . B0 = B Z (R0 ) = 5 kG, as in the previous exercise. Note in passing, . R0 = 50 cm
thus corresponds to . B R = 0.25 T m, . E k = 2.988575 MeV proton kinetic energy,
– radial field index .k = 0 for the time being (constant field at all .(R, θ )),
– a hard-edge field model for .F (Eq. 3.36). In that manner for instance, two
consecutive 60 deg sectors form a continuous 120 deg sector.
A graph of . B Z (R, θ ) can be produced by computing constant radius orbits, for a
series of energies ranging in 0.12–5.52 MeV for instance. DIPOLE[IL = 2] causes
logging of step by step particle data in zgoubi.plt, including particle position and
magnetic field vector; these data can be read and plotted, to yield similar results to
Fig. 3.20.

(b) Concentric trajectories in the median plane.

The optical sequence of Exercise 3.1b (Table 3.4) can be used, by just changing
the INCLUDE to account for a 180.◦ DIPOLE (instead of TOSCA), namely

’INCLUDE’
1
3* 60degSector.inc[#S_60degSectorUnifB:#E_60degSectorUnifB]

wherein 60degSector.inc is the name of the data file of Table 3.6 and
[#S_60degSectorUnifB:#E_60degSectorUnifB]
is the DIPOLE segment as defined in the latter. Note that the segment represents a
60.◦ DIPOLE, thus it is included 3 times.
The additional keywords in that modified version of the Table 3.4 file include
– FIT, which finds the circular orbit for a particular momentum,
– FAISTORE to print out particle data once FIT is completed,
90 3 Classical Cyclotron

Table 3.6 Simulation input data file 60degSector.inc: analytical modeling of a dipole magnet,
using DIPOLE. That file defines the labels (LABEL1 type [16, Sect. 7.7]) #S_60degSectorUnifB
and #E_60degSectorUnifB, for INCLUDEs in subsequent exercises. It also realizes a 60-sample
momentum scan of the cyclotron orbits, from 200 keV to 5 MeV, using REBELOTE

– REBELOTE, which repeats the execution of the sequence (REBELOTE sends

the execution pointer back to the top of the data file) for a new momentum value
which REBELOTE itself defines.
For the rest, follow the same procedure as for Exercise 3.1b. The results are the
same, Fig. 3.21.
3.4 Solutions of Exercises of This Chapter 91

LATERAL EFB

ENTR
OF T
ENTRANCE EFB
θ<0

ANCE
HE M
u
2>
0
>0
AP
FACE
R1

u1
<0
R2>0
θ<0

u 2>0
REFER
ENCE
ω−
<0
TE <0
EXIT EFB
ω +>
0
RM

u1<0

ω− >0
R1<0 <0 u2

R2
θ<0

>0
<0
ACENT u1

AT
R1
>0

EXIT FACE
0 RS OF THE MAP

0
TS >

Fig. 3.24 Parameters used to define the geometry of a dipole magnet with index, using DIPOLE.
In the text, ACENT is noted ACN [16, Fig. 9]

Table 3.7 A.180◦ version of a DIPOLE sector, where the foregoing quantities. AT = 60◦ ,. AC N =
ω+ = −ω− = 30◦ have been changed to . AT = 180◦ , . AC N = ω+ = −ω− = 90◦ —a file used
under the name 180 degSector.inc in further exercises
92 3 Classical Cyclotron

7.63
0.1319 Trev
frev
frev,Trev (non rel.) 7.625
0.1318
7.62
0.1317
7.615

0.1316
7.61
Trev [μs]

frev [MHz]
0.1315 7.605

0.1314 7.6

7.595
0.1313

7.59
0.1312
7.585
0.1311
7.58
10 20 30 40 50 60 70
R [m]

Fig. 3.25 A scan of radius-dependent revolution frequency. An analytical model of a cyclotron

dipole is used, featuring uniform field (no radial gradient, at this point)

(c) Energy and rigidity dependence of orbit radius and time-of-flight.

The orbit radius . R and the revolution time .Trev as a function of kinetic energy . E k
and rigidity . B R are obtained by a similar scan to exercise (b). The procedure is the
same as in Exercise 3.1c. Results are expected to be the same as well (Fig. 3.22).
A comparison of revolution periods can be made using the simulation file of
Table 3.6 which happens to be set for a momentum scan and yields Fig. 3.25, to be
compared to Fig. 3.22: DIPOLE and TOSCA produce the same results as long as both
methods are converged, from the integration step size stand point (small enough),
and regarding TOSCA from field map mesh density stand point in addition (dense
enough).

δR
(d) Numerical convergence: integration step size; . (/\s).
R
This question concerns the dependence of the numerical convergence of the solu-
tion of the differential equation of motion upon integration step size.
Follow the procedure of Exercise 3.1e: a similar outcome to Fig. 3.23 is expected—
ignoring mesh density with the present analytical modeling using DIPOLE.
δR
The . dependence upon the integration step size ./\s is commented in
R
Exercise 3.1e and holds regardless of the field modeling method (field map or ana-
lytical model).

(e) Pros and cons.

Using a field map is a convenient way to account for complicated one-, two- or
three-dimensional field distributions.
3.4 Solutions of Exercises of This Chapter 93

However, using an analytical field model rather, ensures greater accuracy of the
integration method.
CPU-time wise, one or the other method may be faster, depending on the problem.
3.3 Resonant Acceleration

The field map and TOSCA [16, lookup INDEX] model of a 180.◦ sector is used here
(an arbitrary choice, the analytical field modeling DIPOLE would do as well), the
configuration is that of Fig. 3.5 with a pair of sectors.
An accelerating gap between the two dees is simulated using CAVITE[IOPT =
3], PARTICUL is added in the sequence in order to specify ion species and data,
necessary for CAVITE to operate. Acceleration at the gap does not account for the
particle arrival time in the IOPT = 3 option: whatever the later, CAVITE boost will
be the same as longitudinal motion is an unnecessary consideration, here).
The input data file for this simulation is given in Table 3.8. It is resorted to
INCLUDE, twice in order to create a double-gap sequence, using the field map model
of a 180.◦ sector. The INCLUDE inserts the magnet itself, i.e., the #S_halfDipole to
#E_halfDipole TOSCA segment of the sequence of Table 3.3. Note: the theoretical
field model of Table 3.6, segment #S_60degSectorUnifB to #E_60degSectorUnifB
(to be INCLUDEd 3 times, twice), could be used instead: Exercise 3.2 has shown
that both methods, field map and analytical field model, deliver the same results.
Particle data are logged in zgoubi.fai at both occurrences of CAVITE, under the
effect of FAISTORE[LABEL=cavity], Table 3.8. This is necessary in order to access
the evolution of parameters as velocity, time of flight, etc. at each half-turn, given
that each half-turn is performed at a different energy

(a) Accelerate a proton.

A proton with initial kinetic energy .20 keV is launched on its closed orbit radius,
. R0 = p/q B = 4.087013 cm. It accelerates over 25 turns due to the presence to
REBELOTE[NPASS = 24], placed at the end of the sequence. The energy range,
20 keV to 5 MeV, and the acceleration rate: 0.1 MeV per cavity, 0.2 MeV per turn,
determine the number of turns, .NPASS+1 = (5 − 0.02)/0.2 ≈ 25. The accelerated
trajectory spirals out in the fixed magnetic field, it is plotted in Fig. 3.26, reading
data from zgoubi.plt.

(b) Momentum and energy.

Proton momentum . p and total energy . E as a function of kinetic energy, from ray-
tracing (turn-by-turn particle data are read from zgoubi.fai, filled up due to FAIS-
TORE) are √ displayed in Fig. 3.27, together with theoretical expectations, namely,
. p(E k ) = E k (E k + 2M) and . E = E k + M.

(c) Velocity.
Proton normalized velocity .β = v/c as a function of kinetic energy from raytracing
is displayed in Fig. 3.27, together with theoretical expectation, namely, .β(E k ) =
p/(E k + M).
94 3 Classical Cyclotron

Table 3.8 Simulation input data file: accelerating a proton in a double-dee cyclotron, from 20 keV
to 5 MeV, at a rate of 100 kV per gap, independent of RF phase (longitudinal motion is frozen—
see question (e) dealing with CAVITE[IOPT = 7] for unfrozen motion). Note that particle data
are logged in zgoubi.fai (under the effect of FAISTORE) at both occurrences of CAVITE. The
INCLUDE file FieldMapSector.inc is taken from Table 3.3
3.4 Solutions of Exercises of This Chapter 95

0.8
"zgoubi.plt" u ($42==noel_1? $22:$22+pi ):($10 *cm2m)
"zgoubi.plt" u ($42==noel_1? $22:$22+pi ):($10 *cm2m)
0.6

0.4

0.2
YLab [m]

0
0 0.2 0.4 0.6 0.8

-0.2

-0.4

-0.6

-0.8
-0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8
XLab [m]

Fig. 3.26 Twenty five turn spiral trajectory of a proton accelerated in a uniform 0.5 T field from
20 keV to 5 MeV at a rate of 200 kV per turn (a 100 kV gap voltage). The vertical thick line materi-
alizes the gap, the upper half (red) corresponds to the first occurrence of CAVITE in the sequence
(Table 3.8), the lower half (blue) corresponds to the second occurrence of CAVITE

100 943.5 0.11

p(Ek) E(Ek) β(Ek)
theor. theor theor.
943
90 0.1

942.5
0.09
80
942
0.08
70 941.5
E [MeV/c]
p [MeV/c]

941 0.07
v/c

60
940.5 0.06

50 940
0.05
939.5
40
0.04
939
30 0.03
938.5

20 938 0.02
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5
Ek [MeV] Ek [MeV]

Fig. 3.27 Energy dependence of, left: proton momentum. p (left axis) and total energy. E (right axis)
and of, right: proton normalized velocity.β = v/c. Markers: from raytracing; solid lines: theoretical
expectation

(d) Relative velocity, orbit length and time of flight.

The relative increase in velocity is smaller than the relative increase in orbit length
as energy increases (this is what Fig. 3.28 shows). Thus the relative variation of the
revolution time, Eq. 3.23, is positive; in other words the revolution time increases
with energy, the revolution frequency decreases. Raytracing outcomes are displayed
in Fig. 3.28, they are obtained using the gnuplot script given in Table 3.8. Note that
the path length difference (taken as the difference of homologous quantities in a
96 3 Classical Cyclotron

ΔC/C
dβ/β
ΔT/T=dC/C-dβ/β
theor. ΔT/T
1
Δβ/β, ΔC/C, ΔTrev/Trev

0.1

0.01
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Ek [MeV]

Fig. 3.28 Relative variation of velocity ./\β/β (empty circles), circumference ./\C/C (solid disks)
and revolution time ./\T /T (triangles), as a function of energy, from raytracing. Theoretical expec-
tation for the latter is also displayed (solid line), for comparison

common line) is always between the two CAVITEs (particle data are logged at the
two occurrences of CAVITE), crossed successively, which is half a turn. Same for
the difference between homologous velocity data on a common line, it corresponds
to two successive crossings of CAVITE, i.e., half a turn. The graph includes the
theoretical .δTrev /Trev (Eq. 3.23) for comparison with raytracing; some difference
appears in the low velocity regime, this may be due to the large ./\β step imparted
by the 100 kV acceleration at the gaps.

(e) Harmonic h = 3 RF.

The input data file for this simulation is given in Table 3.9. The RF is on harmonic
h = 3 of the revolution frequency. It has been tuned to ensure acceleration up to 3 MeV.
The accelerating gap between the two dees is simulated using CAVITE[IOPT = 7]:
by contrast with the previous exercise (where CAVITE[IOPT = 3] is used), the RF
phase at ion arrival at the gap is now accounted for.
Repeating questions (b–d) is straightforward, changing what needs be changed in
Table 3.9 input data file.

3.4 Spin Dance

The DIPOLE analytical field model of Exercise 3.2 (Table 3.6) is used here, as
opposed to using a field map and TOSCA, as it allows more straightforward changes
in the field, if desired.
3.4 Solutions of Exercises of This Chapter 97

Table 3.9 Simulation input data file: accelerating a proton in a double-dee cyclotron, from 20 keV
to 5 MeV, using harmonic 3 RF. The INCLUDE file is taken from Table 3.6

(a) Spin transport.

Spin transport is obtained by adding SPNTRK. PARTICUL is necessary in order to
get the Thomas-BMT equation of motion solved [16, Sect. 2]. This results in the input
data file given in Table 3.10 (excluding FIT and REBELOTE keywords, introduced
for the purpose of the following question (b)).

The use of SPNTRK results in the following outcome (an excerpt from zgoubi.res
execution listing):

Spin coordinates are logged in zgoubi.res execution listing using SPNPRT. Five
sample passes around the cyclotron (four iterations by REBELOTE) result in the
following outcomes in zgoubi.res, under SPNPRT:
98 3 Classical Cyclotron

Table 3.10 Simulation input data file: add spin to the cyclotron simulation of Table 3.6. The present
input file INCLUDEs six copies of the 60.◦ sector DIPOLE defined therein

(b) Spin precession.

Proton case is considered, simulation is performed using Table 3.10 input data file.
Initial spin is parallel to the X axis (longitudinal). The particle is raytraced on the
circular closed orbit over one revolution, for a particular momentum. Particle data
resulting from a FIT (FIT forces orbit closure, by varying the initial .Y0 ) are logged
in zgoubi.fai, by FAISTORE. The computation is repeated using REBELOTE in the
very manner that the energy scan was done in Exercise 3.2, over an energy range
12 keV.→5 MeV.
Figure 3.29 (obtained using the gnuplot script given in Table 3.10) displays the
resulting energy dependence of the spin precession, .θsp (E), together with its differ-
ence to theoretical expected .θsp (E) = G ME
× 2π = Gγ × 2π (proton gyromagnetic
anomaly .G = 1.792847).
3.4 Solutions of Exercises of This Chapter 99

1.803 0.01000000
θsp/2π

1.802 1 cm
5 mm
0.00100000
Spin precession angle θsp / 2π 1.801

1.8

relative difference num./theor

0.00010000

1.799

1.798 0.00001000

1.797

0.00000100
1.796

1.795
0.00000010

1.794

1.793 0.00000001
1.793 1.794 1.795 1.796 1.797 1.798 1.799 1.8 1.801 1.802 1.803
Gγ

Fig. 3.29 .Gγ dependence of the spin precession angle over a revolution around the cyclotron,
in the moving frame (left axis), and relative difference to .Gγ for the two integration step sizes
./\s = 0.5 and 1 cm (right axis), Markers are from raytracing, solid lines are to guide the eye

(c) Spin tune.

Two protons are injected with longitudinal initial spin .Si || OX axis and respective
energies 12 keV and 5.52 MeV, thus the following OBJET (a slight modification to
Table 3.10 data):

FAISCEAU following FIT (Table 3.10) allows to control that momentum and
trajectory radius are matched, which means coordinates at OBJET and current coor-
dinates at FAISCEAU are equal. Inspection of zgoubi.res execution listing shows for
instance, after 4 turns:

A graphic of the projection of the spin motion on the longitudinal axis, over a
few turns, from the ray tracing, is given in Fig. 3.30, together with the longitudinal
component as of the parametric equations of motion
{
S X = Ŝ cos(Gγ θ )
. (3.37)
SY = Ŝ sin(Gγ θ )
100 3 Classical Cyclotron

Fig. 3.30 Longitudinal spin component motion (left vertical axis), observed in the moving frame,
case of 0.2 MeV energy, R = 12.924888 cm (left graph), and of 5.52 MeV energy, R = 67.998 cm
(right graph). Markers are from ray tracing, the solid line is the theoretical expectation (Eq. 3.37).
The right vertical axis (triangle markers; solid line is to guide the eye) shows the absolute difference
between both. The oscillation is as expected slightly faster at 5.52 MeV: frequencies are in the ratio
.γ (5.52 MeV)/γ (0.2 MeV) = 1.00566

The motion amplitude is . Ŝ = sin φ, with .φ the angle that the spin vector makes with
the vertical precession axis. In this simulation .S is launched parallel to OX, thus
.φ = π/2 and . Ŝ = 1.

Now, checking the spin precession:

Placing both FAISCEAU and SPNPRT commands right after the first dipole sec-
tor allows checking the spin precession and its relationship to particle rotation, for
simplicity right after the first pass through that first sector, as follows. FAISCEAU
and SPNPRT (Table 3.10) yield, respectively:

SPNPRT tells that,

– case of the first particle, tagged ‘m’ above; its energy is 200 keV, .γ = 1.00021315,
its spin tune is .νsp = Gγ = 1.793229
The computed value of the ‘.(Si , S f )’ angle between initial and final spin vectors
is –107.594 (truncated), negative as spin precession has the sign of proton rotation.
Theoretical expectation is .Gγ α = −107.59377 deg. The resulting spin components
are, as above, . S X = cos(−107.59377) = −0.302266 and . SY = sin(−107.59377) =
−0.9532235.
– case of the second particle, tagged ‘o’; its energy is 5.52 MeV, .γ = 1.00588315,
its spin tune is .νsp = Gγ = 1.803394
The computed value of ‘.(Si , S f )’ is –108.204 (truncated). Theoretical expectation is
. Gγ α = −108.20370 deg.
3.4 Solutions of Exercises of This Chapter 101

Now, accounting for particle rotation in order to get spin coordinates in the labo-
ratory frame:
– the FAISCEAU outcome above shows that, after crossing the 60 deg sector the
angles of the two particles have the value .T = 0, which is expected as they are
launched with zero incidence, and as DIPOLE uses a polar coordinate system [16]
with particle coordinates computed in the moving (rotating) frame. The latter has
also undergone a -60 deg rotation, clockwise, which is therefore the implicit rotation
of the particles in the laboratory frame. The spin precession in the laboratory frame
results, namely,
– case of the first particle: .(1 + Gγ )α = −167.59377 deg.
– case of the second particle: .(1 + Gγ )α = −168.20370 deg.

(d) Spin dance.

A 200 keV proton is injected with its initial spin vector at 80 degrees from the vertical
axis. The input data file for this simulation is given in Table 3.11, together with a
gnuplot script for the animation. The latter plots three things, concurrently:
– the circular trajectory of the particle in the (X,Y) plane: this is the curve at Z = 0
in Fig. 3.31, a set of points .{(R cos(−X ), R sin(−X ), 0)} resulting from the step by
step integration. Note that X is counted positive clockwise in zgoubi.fai (consistently
with the definition of DIPOLE parameters, Fig. 9 in [16]), hence “–X” the rotation
angle;
– the spin vector: its foot is attached to the particle (the previous set of points),
whereas its tip is at .{(S X cos(−X ) − SY sin(−X ), S X sin(−X ) + SY cos(−X ), S Z },
with . S X , SY , S Z the spin vector components in the moving frame as read from
zgoubi.fai.
( . S Z is constant as the precession axis is parallel to the . Z axis. The
)
cos(−X ) − sin(−X )
. rotation applied to the .(S X , SY ) vector accounts for the
sin(−X ) cos(−X )
transformation from the moving frame to the laboratory frame;
– the cycloidal shape trajectory of the tip of the spin vector (the previous set of
points).
A frozen view of that spin dance, over about 2.5 proton revolutions around the
ring, is given in Fig. 3.31.

(e) Deuteron
The input data file set up for questions (b–e) can be used mutatis mutandis, as follows.
Raytracing a different particle requires changing the reference rigidity, BORO,
under OBJET, and changing particle data, under PARTICUL. That reference rigidity
is to be determined from the field value in the dipole model (namely, . B0 = 5 kG).
102 3 Classical Cyclotron

Table 3.11 Simulation input data file: spin dance, 20 turns around a uniform field cyclotron. The
INCLUDE file 60degSector.inc is taken from Table 3.6

Particle data for these two particles are (respectively mass (MeV/c.2 ), charge (C),
G factor):

.deuter on : 1875.612928 1.602176487 × 10−19 − 0.14301

3
. H e2+ : 2808.391585 3.204352974 × 10−19 − 4.1841
3.4 Solutions of Exercises of This Chapter 103

Fig. 3.31 Dance—frozen, here—of the spin of a 200 keV proton over 2.5 turns around the cyclotron.
The circle on the left, or bottom closed curve on the right, is the trajectory of the proton. The cycloidal
curve represents the motion of the spin vector tip in the moving frame

3.5 Synchronized Spin Torque

The simulation input data file of Exercise 3.4(d) can be used here, with a few addenda
or modifications, as follows:
(i) the initial ion coordinate D (rigidity relative to the reference
BORO = 64.6244440) under OBJET has to be calculated for the four energies con-
cerned;
(ii) the closed orbit radius at 0.2, 108.412, 118.878 and 160.746 MeV has to be
found; calculation is straightforward given that the field considered here is vertical,
uniform, namely, . B Z = constant = 5 kG, .∀R, so that . R = Bρ/B Z ; otherwise a FIT
procedure can be used to find the orbit radius, given the rigidity, as done already
in various exercises [16, lookup “closed orbit”], that could help for instance in the
presence of a radial index, or field defects;
(iii) initial spins are set vertical for convenience, but this is not mandatory;
(iv) the multiturn tracking is set to a few tens of turns, in order to allow a few spin
precessions;
(v) particle data through DIPOLEs are saved step-by-step all the way in zgoubi.plt
by means of IL = 2 (the integration step size is 1 cm (Table 3.6), thus zgoubi.plt may
end up bulky);
(vi) turn-by-turn data are saved in zgoubi.fai by means of FAISTORE;
(vii) SPINR is added at the end of the sequence, to impart on spins the requested
X-tilt.
This results in the updated simulation input data file given in Table 3.12.

The oscillatory motion of the vertical spin component as the ion orbits around the
ring, is displayed in Fig. 3.32. The spin points upward, parallel to the vertical axis at
start; SPINR kick is 10 deg in the present case. At .Gγ = 2 the spin always finds itself
back in the (Y, Z) transverse plane after one proton orbit, this synchronism causes
the cumulated spin tilt at SPINR to take the value . N × 10 deg (with N the number of
orbits). Thus after 18 proton orbits, 36 spin precessions, the spin points downward;
104 3 Classical Cyclotron

Table 3.12 Simulation input data file: superimposition of a turn-by-turn localized.10 deg X-rotation
of the spin (using SPINR[.φ = 0, μ = 10]), on top of Thomas-BMT .2π Gγ Z-precession. The
INCLUDE file 60degSector.inc is taken from Table 3.6

Fig. 3.32 . S Z motion versus orbital angle, while the ion orbits on a circle. . S Z is constant over a
turn and then undergoes a discontinuity upon the 10 deg X-tilt, hence the step function. At .Gγ = 2
it takes 36 turns, or 226.194 rad, to complete an oscillation. A graph obtained using zpop: menu 7;
1/1 to open zgoubi.plt; 2/[6,23] for . S Z versus .θ ; 7 to plot

it takes 36 orbits, or 226.194 rad, to complete an oscillation. If .Gγ moves away from
an integer, the spin tilts with bounded amplitude, within the limits of a cone.
Additional graphs and details are obtained using the simulation file of Table 3.13.
This file simulates spin motion in three different cases, .Gγ = 1.79322, .Gγ = 2,
integer, yielding an integer number of spin precessions over one proton orbit around
3.4 Solutions of Exercises of This Chapter 105

Table 3.13 Simulation input data file: a similar simulation to Table 3.12, for different .Gγ values,
namely 1.79322, 2 and 2.5. The spin kick at SPINR has been changed to 20 deg. Regarding the use
of OBJET[IEX] option: IEX = –9 allows inhibiting the tracking for the particle(s) concerned, all
the rest left unchanged; it is necessary here to have at least one particle with IEX = 1, for proper
operation of the gnuplot scripts. The INCLUDE file 60degSector.inc is taken from Table 3.6
106 3 Classical Cyclotron

Fig. 3.33 Top row: spin coordinates versus turn; middle row: projection in the median plane (the
segment between two consecutive circles materializes the location of the X-kick by SPINR); bottom
row: projection on a sphere..Gγ = 1.793229: far from an integer,.S remains within a cone of reduced
aperture. .Gγ = 2: the spin vector oscillates between up and down orientations, by 20 deg steps; it
takes 180/20 = 9 orbits for the X-precession at SPINR to flip the spin; .Gγ = 2.5: the spin vector
finds itself back in the (Y,Z) plane at the location of SPINR, after one orbit and a half-integer number
of precessions; it alternates between vertical and 20 deg from vertical, after each orbit around the
cyclotron

the cyclotron, and .Gγ = 2.5, half-integer, yielding a half-integer number of spin
precessions over one proton orbit. Outcomes are given in Fig. 3.33 which shows the
spin motion projected on the (X,Y) plane (horizontal), and on a sphere, step-by-step.
The spin kick by SPINR is 20 deg in this case. If.Gγ = 1.793229, far from an integer,
.S, initially vertical, remains at a bounded angle to the vertical axis, X-kicked from
one circle to another, turn after turn; if .Gγ = 2 the spin vector flips by 20.◦ in the
(Y, Z) plane at SPINR, turn after turn; if .Gγ = 2.5, half-integer, the spin vector
undergoes a half-integer number of precessions over one orbit around the cyclotron,
it jumps and alternates between vertical, and the surface of the 20.◦ Z-axis cone.
3.4 Solutions of Exercises of This Chapter 107

3.6 Weak Focusing

(a) Add a field index.

To the first order in . R, in the median plane (Z = 0) and noting . R = R0 + d| R,
∂B |
. B Z (R0 ) = B0 , . B Z (R) = B, the field writes (Sect. 3.2.2) . B(R) = B0 + d R .
∂ R R0
R0 ∂ B
With .k = B0 ∂ R (Eq. 3.11) this yields

B0
. B(R) = B0 + k dR (3.38)
R0

Assume the earlier 200 keV conditions as a reference, thus take

. R0 = 12.9248888 cm as the 200 keV radius, whereas B0 = B(R0 ) = 5 kG.

Take.k = −0.03, a slow decrease of the field with. R—proper to ensure appropriate
vertical focusing with marginal impact on the radial extent of the cyclotron. For
instance, with that index value the 5 MeV orbit is at a radius of .75.75467 cm (see
OBJET in Table 3.3) (giving . B = 0.3235 T along the orbit), whereas if k = 0 then
. R = 75.75467 cm is the 6.8463 MeV orbit radius (. B = 0.3788 T).
The field map is generated using a similar Fortran program to that of Exercise 3.1
(see Table 3.1), mutatis mutandis, namely, introducing a reference radius . R0 and
field index .k. The resulting program is given in Table 3.14, it can be compiled and

Table 3.14 A Fortran program which generates a.60◦ mid-plane field map with non-zero transverse
field .k. The field map it produces is logged in geneSectorMapIndex.out
108 3 Classical Cyclotron

Table 3.15 First and last few lines of the field map file geneSectorMapIndex.out. The file starts
with an 8-line header, the first one of which is effectively used by zgoubi, the following 7 are just
comments

Fig. 3.34 Left: field map of a 60 deg magnetic sector with radial index, 76 cm radial extent. The
field decreases from the center of the ring (at .(X Lab , YLab ) = (0, 0)). Right: three circular arc of
trajectories over a sextant, at respectively from left to right: 0.02 MeV, 0.2 MeV (energy on the
reference radius) and 5 MeV

executed, as is, excerpts of the field data file so obtained are given in Table 3.15, a
graph . B Z (R, θ ) is given in Fig. 3.34. The orbit radius is assessed for three different
energies, and appears to be in accord with theoretical expectation (Fig. 3.34-right).
Comparison with Fig. 3.20-right shows the effect of the negative index on the radial
distribution of the orbits, including a radius about 20% greater in the 5 MeV range.
The input data file to find these trajectories is given in Table 3.16:
3.4 Solutions of Exercises of This Chapter 109

Table 3.16 Simulation input data file FieldMapSectorIndex.inc: a file to test trajectories for a
field map with radial index. This file also defines the INCLUDE segment between the LABEL_1s
#S_60degSectorIndx and #E_60degSectorIndx

– the file defines an INCLUDE segment, #S_60degSectorIndx to #E_60degSector

Indx, used in subsequent exercises;
– the file is set to allow a preliminary test regarding the field map geneSectorMapIn-
dex.out (as produced by the program given in Table 3.14), by computing three circular
trajectories centered on the center of the map, at respectively 20 keV, 200 keV (the
reference energy for the definition of the gradient index.k) and 5 MeV (a large radius);
– note that once the FIT procedure is completed, zgoubi continues in sequence, so
raytracing the 3 ions through the field map with, this time, IL set to 2 under TOSCA
for stepwise particle data to be logged in zgoubi.plt.
110 3 Classical Cyclotron

(b) R-dependence of orbit rigidity.

The method is similar to Exercise 3.1(b) (see Table 3.4): FIT finds the closed orbit
radius . R for a given ion rigidity, and REBELOTE is used to repeat for a series of
different momenta, 20 here. The input data file for this exercise is given in Table 3.17,
it includes a 21 ion 1-turn raytracing, in sequence with the previous 21-orbit finding.

Table 3.17 Simulation input data file: scan orbits for momentum dependence. Two problems are
stacked, executed in sequence: in a first stage FIT finds a closed orbit, whose coordinates are logged
in initialRs.fai file when FIT is completed, following what REBELOTE repeats for an additional 20
momenta; in a second stage OBJET grabs the 21-set of ion coordinates from initialRs.fai and these
ions are raytraced over 6 sectors, i.e., one full turn. The INCLUDE file FieldMapSectorIndex.inc
is taken from Table 3.16
3.4 Solutions of Exercises of This Chapter 111

0.8 0.35 0.35

theor.

0.6
0.3 0.3

0.4
0.25 0.25

0.2
0.2 0.2
YLab [m]

Bρ [T m]
0
0 0.2 0.4 0.6 0.8
0.15 0.15
-0.2

0.1 0.1
-0.4

0.05 0.05
-0.6

-0.8 0 0
-0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
XLab [m] R [m]

Fig. 3.35 Case of field index k = –0.03. Left: closed orbits at a series of different rigidities. Right:
comparison of . Bρ(R) from raytracing outcomes (markers) and from theory (solid line, Eq. 3.39)

Table 3.18 Simulation input data file sectorWithIndex.inc: definition of a dipole with index, case
of analytical field modeling, namely here k = –0.03 and reference radius . R0 = 50 cm. Definition of
the [#S_60degSectorWIdx:#E_60degSectorWIdx] segment

Raytracing outcomes for .k = −0.03, . R0 = R(E = 200 keV) = 12.924888 cm,

. B0 = B(R0 ) = 0.5 T are given in Fig. 3.35, together with theoretical expectation
(with B(R) from Eq. 3.7)
( )
R − R0
Rigidity B R(R) = B0 1 +
. k R (3.39)
R0

(c) Paraxial motion.

A proton with energy 1 MeV is considered, here. DIPOLE [16, lookup INDEX] is
used rather than a field map, so to allow to freely change the .k index value (using
TOSCA instead would require computing a new field map when changing .k).
The input data for a 60 deg sector are given in Table 3.18, essentially a copy of
the uniform dipole field case of Table 3.6 in which the index value .k = −0.03 has
been added (line 3 under DIPOLE). The input data sequence for multiturn trajectory
112 3 Classical Cyclotron

Table 3.19 Simulation input data file: scan orbits for momentum dependence; the file actually
stacks two simulations, executed in sequence; the second simulation uses data produced by the
first one, as follows. The first part of the file finds the closed orbits, they depend on the vertical
excursion and are not exactly zero, due to the field index; closed orbit coordinates so found are logged
in initialRs.fai when FIT is completed. The second part of the file starts at the second occurrence of
OBJET which reads initial particle coordinates from initialRs.fai and tracks these particles through
a sequence of 120 sector dipoles, i.e., 20 turns. The [#S_60degSectorWIdx:#E_60degSectorWIdx]
segment of Table 3.18 is INCLUDEd, here

computation around the cyclotron is given in Table 3.19: in a first stage, orbit finding
is performed by FIT, for 1 MeV energy; in a subsequent second stage, 4 protons with
their initial horizontal coordinates taken on the closed orbit, and differing by their
initial vertical take-off angle, are tracked over 120 sectors, i.e., 20 turns around the
ring.
3.4 Solutions of Exercises of This Chapter 113

0.0008 0.0003
P[mrad]=0.3 P[mrad]=0.3
P[mrad]=0.2 P[mrad]=0.2
P[mrad]=0.1 P[mrad]=0.1
0.0006 P[mrad]=0.0 P[mrad]=0.0
0.0002

0.0004

0.0001
0.0002
Z [m]

Z [m]
0 0

-0.0002
-0.0001

-0.0004

-0.0002
-0.0006

-0.0008 -0.0003
0 5 10 15 20 25 30 35 0 5 10 15 20 25 30 35
s [m] s [m]

Fig. 3.36 Vertical sine motion over a few turns around the cyclotron,
√ at 1 MeV. Vertical take-off
angles are . P0 = 0, 0.1, 0.2,
√ 0.3 mrad. Left: k = –0.03, .ν Z = 0.03 ≈ 0.173 oscillations per turn;
right: for k = –0.2, .ν Z = 0.2 ≈ 0.447 oscillations per turn
0.28539608
P[mrad]=0.3 P[mrad]=0.3
P[mrad]=0.2 0.26415100 P[mrad]=0.2
P[mrad]=0.1 P[mrad]=0.1
P[mrad]=0.0 P[mrad]=0.0
0.28539607

0.26415100
0.28539607

0.26415099
0.28539606
Y [m]

Y [m]

0.26415099
0.28539606

0.26415098
0.28539605

0.28539605 0.26415098

0.28539604 0.26415097
0 5 10 15 20 25 30 35 0 5 10 15 20 25 30 35
s [m] s [m]

Fig. 3.37 Horizontal motion at 1 MeV, 20 turns around

√ the cyclotron, for vertical take-off angles
. P0= 0, 0.1, 0.2, 0.3
√ mrad. Left: k = –0.03, .ν R = 1 + 0.03 ≈ 1.015 oscillations per turn; right:
for k = –0.2, .ν R = 1 + 0.2 ≈ 1.095 oscillations per turn

Figure 3.36 displays the vertical sine motion. Stronger index (.k closer to –1) results
in stronger vertical focusing, hence more oscillations as expected from Eq. 3.18 and
smaller motion amplitude as expected from Eq. 3.17. The latter can be written
√
R0 −k R0
. Z (s) = P0 √ sin (s − s0 ) and Ẑ = P0 √ (3.40)
−k R0 −k

Note that this vertical oscillation results in a modulation of the field along the
trajectory (see question (d) of this exercise) which results in a radial oscillation, a
second order Y–Z coupling effect (extremely weak), displayed in Fig. 3.37.

(d) Magnetic field.

The magnetic field experienced by 1 MeV protons with four different take-off angles
. P0 (Fig. 3.36), along their respective trajectories, case of an index value .k = −0.03,
is displayed in Fig. 3.38. It is essentially constant as expected.
114 3 Classical Cyclotron

Fig. 3.38 Magnetic field 0.50643814

P[mrad]=0.3
P[mrad]=0.2
experienced by 1 MeV P[mrad]=0.1
P[mrad]=0.0
protons with four different 0.50643814

take-off angles . P0
(Fig. 3.36), along their 0.50643813
respective trajectories. Case
.k = −0.03. The stepwise

BZ [T]
structure of these . B Z (s) 0.50643813

curves is due to the fact that

field variations are at the 0.50643812

limit of computer truncation

related accuracy 0.50643812

0.50643812
0 5 10 15 20 25 30 35 40
s [m]

3.7 Loss of Isochronism

In order to scan .Trev (R) for different .k values, DIPOLE [16, lookup INDEX] is used
here, as it allows to easily vary .k and subsequently find the closed orbit using FIT.
The method of Exercise 3.6 is employed to obtain a scan. The input data file of
Table 3.17 is a good starting point to do this exercise, changing the INCLUDE to
account for DIPOLE instead of a field map modeling using TOSCA: the proper
INCLUDE formatting can be reproduced from Table 3.19. IL under DIPOLE may
be set at IL = 0 as zgoubi.plt is not used here. Introduce FAISTORE to store local
particle data after FIT (that includes time of flight, the quantity of interest here, which
requires PARTICUL[PROTON] following OBJET).
The new input data file so built for this simulation, is given in Table 3.20.

⎧ DIPOLE (cf.
This input data file is run for four different .k values, namely, under
⎨ 30. 5. 0 0. 0.
Table 3.18), the line “30. 5. –0.03 0. 0.” is successively changed to .⎩ 30. 5. − 0.5 0. 0. .
30. 5. − 0.95 0. 0.
The corresponding zgoubi.fai files are saved under dedicated copies for plotting, see
“gnuplot script gnuplot_Zfai_scanTrev.gnu” at the bottom of Table 3.20.
The results of these .Trev scans are displayed in Fig. 3.39. In the case .k = 0 the
loss of isochronism is only due to the relativistic change of the mass, a non-zero k
augments the effect. The loss of isochronism is the cause of the .≈ 20 MeV proton
energy limit of the classical cyclotron.

3.8 Ion Trajectories

A zgoubi data file is set up for computation of particle trajectories, taking a field value
on reference radius of. B0 (R0 ) = 0.5 T, and reference energy 200 keV (proton). These
hypotheses determine the reference radius value. DIPOLE [16, lookup INDEX] is
used (Table 3.21), for its greater flexibility in changing magnet parameters, field and
radial field index amongst other, compared to using TOSCA and a field map.
3.4 Solutions of Exercises of This Chapter 115

Table 3.20 Simulation input data file: scan revolution time. The [#S_60degSector
WIdx:#E_60degSectorWIdx] segment of Table 3.18 is INCLUDEd, here

0.14 0.13134 0.134

0.13 0.13132
0.133

0.12 0.1313
0.132

0.11 0.13128
Trev at k=0[μs]

0.131
k=0
Trev [μs]

Trev [μs]

0.1 k=-0.03 0.13126 k=0

k=-0.5 k=-0.03
k=-0.95
0.13
0.09 0.13124

0.129
0.08 0.13122

0.128
0.07 0.1312

0.06 0.13118 0.127

0 0.05 0.1 0.15 0.2 0.25 0.3 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
R [m] R [m]

Fig. 3.39 A scan of the revolution time, from 0.02 to 1 MeV, and its dependence on the field index
.k.The right vertical axis only concerns the case .k = 0 where the change in revolution time is weak
and only due to the mass increase (in .Trev = 2π γ m 0 /q B). The right graph shows, up to 5 MeV,
the relatively important contribution of the focusing index, even a weak k = –0.03, compared to the
effect of the mass increase (k = 0 curve). Markers are from raytracing, solid lines are from theory
116 3 Classical Cyclotron

Table 3.21 Input data file 60DegSectorR200.inc: it defines DIPOLE as a sequence seg-
ment comprised between the “LABEL_1” type labels [16, Sect. 7.7] #S_60DegSectorR200 and
#E_60DegSectorR200. DIPOLE here, has an index .k = −0.03, reference radius . R0 ≡ R0 (E k =
200 keV) = 12.924888 cm and . B0 = B(R0 ) = 0.5 T. Note that (i) this file can be run on its own: it
has been designed to provide the transport MATRIX of that DIPOLE; (ii) for the purpose of some
of the exercises, IL = 2 under DIPOLE, optional, causes the printout of particle data in zgoubi.plt,
at each integration step (this is at the expense of CPU time, and memory volume)

(a) Transverse motion.

It first has to be checked that there is consistency between initial orbital radius .Y0
in OBJET at 200 keV proton energy and the value of the reference radius . R0 in
DIPOLE (Eq. 3.35). Its theoretical value is . R0 = B O R O/5[kG] = 12.924889 cm,
a closed orbit finding using FIT can be performed, or it can be referred to the solutions
of earlier exercises, to check agreement with raytracing outcomes.

(b) Wave numbers at 1 and 5 MeV.

These considerations result in the input data file given in Table 3.22, to compute
multiturn trajectories.; note that . R0 = 12.924889 cm therein, whereas a value of
, ,,
. R0 = 50 cm may be taken instead in other exercises. Field index derivatives.k , k , ...
are taken null in the present exercise.
Three particles with paraxial radial and axial motions are raytraced over a few
turns. Their starting radius is the closed orbit radius for the respective energies, while
a 0.1 mrad take-off angle is imparted to each particle both vertically and horizontally.
The value of the focusing index .k E at an energy . E can be expressed in terms of
DIPOLE data which are, the index value .k at . R0 (Eq. 3.11), reference radius . R0 , and
field . B0 = B Z (R0 ), namely,

RE ∂ B R0 + /\R ∂ B 1 + /\R/R0 [ /\R ]

k =
. E = ≈k ≈ k 1 + (1 − k)
BE ∂ R B0 + /\B ∂ R 1 + k/\R/R0 R0
3.4 Solutions of Exercises of This Chapter 117

Table 3.22 Simulation input data file: raytrace a few turns around the cyclotron, three particles with
different momenta, and 0.1 mrad horizontal and vertical take-off angles. The INCLUDE segment
is taken from Table 3.21

with ./\R assumed small, .∂ B/∂ R = k B E /R E an energy independent quantity, and

the index E denoting a quantity taken at the reference energy. The latter property is
illustrated in Fig. 3.40, produced using the input data file of Table 3.23.
The resulting radial and axial motions over 10 turns are displayed in Fig. 3.41,
which also illustrates, for paraxial motion at some reference energy, the energy
dependence of the focusing strength (or wave number) and of the motion amplitude.
An estimate of the wave numbers can be obtained as the inverse of the number of
turns per oscillation, namely,
| |
C E || C E ||
.ν R = and ν Z =
/\s M | E /\s M | E

with ./\s M the measured distance between two consecutive maxima in the sinusoid
of concern in Fig. 3.41, .C E the closed orbit length for the energy of concern. Both
quantities are obtained from motion records in zgoubi.plt. This yields the values
of Table 3.24, where √ they are compared √ with the theoretical expectations, namely
(Eq. 3.18), .ν R = 1 + k and .ν Z = −k.
The maximum amplitude of the oscillation is obtained from zgoubi.plt records
as well, this yields the results of Table 3.25. For comparison, the theoretical values
are (Eqs. 3.16 and 3.17 with respectively .x0 = 0, x0, = T0 and . y0 = 0, y0, = P0 )
.Ŷ = T0 √ and . Ẑ = P0 √R−k
RE
1+k
E
. wherein . R E denotes the closed orbit radius at energy
E (for the record: . R E ≡ R0 at energy . E = 200 keV, in the foregoing).
118 3 Classical Cyclotron

1 2 3 4 5 6 7 8
5.1

4.9

4.8
BZ [kG]

4.7

4.6

4.5

4.4

4.3

4.2
1 2 3 4 5 6 7 8
R [m]

Fig. 3.40 In DIPOLE field model (Eq. 3.35), . ∂∂ BR is constant: this graph shows the linear decrease
of the field . B Z (R) (Eq. 3.38), obtained from the raytracing of particles circulating in the median
plane on orbits spanning a 0.2–5 MeV energy range

(c) Comparison with theory.

Figure 3.42 shows the difference between numerical and theoretical vertical motion
excursion, using an ad hoc gnuplot script. An integration step size ./\s = 2 cm is used
in the numerical integration.

(d) A scan of energy dependence of wave numbers.

A scan of the wave numbers over 200 keV.−5 MeV energy range, computing tunes
with MATRIX, is performed using the input data file given in Table 3.26 (essentially
a copy of the input data file of Table 3.23, with an INCLUDE accounting for 6
DIPOLEs [16, lookup INDEX]).
OBJET[KOBJ = 5] generates 13 particles with paraxial horizontal, vertical and
longitudinal sampling, proper to allow the computation of the first order transport
coefficients and wave numbers by MATRIX. REBELOTE repeats MATRIX compu-
tation for a series of different particle rigidities. It is preceded by FIT which finds
the closed orbit. MATRIX includes a PRINT command, which causes the transport
coefficients (and various other outcomes of MATRIX computation) to be logged
in zgoubi.MATRIX.out. This allows producing the graphic in Fig. 3.43—using the
gnuplot script given at the bottom of Table 3.26.
3.4 Solutions of Exercises of This Chapter 119

Table 3.23 Simulation input data file for a magnetic field scan. The INCLUDE segment is taken
from Table 3.21

0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10
0.0001 0.02

-5
8x10
0.015
-5
6x10
0.01
4x10-5

-5 0.005
2x10
Y [cm]

Z [cm]

0 0

-5
-2x10
-0.005

-4x10-5
-0.01
-5
-6x10

-5 -0.015
-8x10

-0.0001 -0.02
0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10
s /C E s /C E

Fig. 3.41 Radial (left) and axial (right) paraxial motion around respectively the 200 keV (smallest
amplitude), 1 MeV (intermediate) and 5 MeV (greatest amplitude) closed orbit (the latter is circular,
in the median plane, with radius respectively . R200 keV = 12.924888 cm, . R1 MeV = 30.107898 cm
and . R5 MeV = 75.754671 cm). The horizontal axis in this graph is .s/C E : path length over closed
orbit circumference at energy E, the vertical axis is the motion excursion
120 3 Classical Cyclotron

Table 3.24 Wave numbers, from numerical raytracing (columns denoted “ray-tr.”), from theory,
and from discrete Fourier transform (‘DFT’ cols.) from a multi-turn tracking

E (MeV) .k E .ν R = .ν Z=
√ √
ray-tr. . 1+k DFT ray-tr. .−k DFT
0.2 –0.03 0.98520 0.9849 0.98513 0.17320 0.1732 0.17321
1 –0.07279 0.96187 0.96292 0.96291 0.26980 0.26979 0.26981
5 –0.20586 0.89083 0.89115 0.89115 0.45326 0.45371 0.45371

Table 3.25 Maximum amplitude of the oscillation, from raytracing (columns denoted “ray-tr.”)
and from theory. . R E is the closed orbit radius for the energy of concern, .T0 = P0 = 0.1 mrad is the
trajectory angle at the origin, positions at the origin are zero
E (MeV) k .Ŷ . Ẑ
R R
ray-tr. . T0 √ E ray-tr. . P0 √ E
1+k −k
(.×10−5 ) (.×10−5 )
0.2 –0.03 1.3123 1.3125 7.4622 7.4624
1 –0.072787 3.1270 3.1267 1.1160 1.1160
5 –0.20586 8.5010 8.5008 1.6697 1.6697

Fig. 3.42 Vertical excursion 0 1 2 3 4 5 6 7 8 9 10

0.000010000
z(x)
of a 1 MeV trajectory over 20 0.1

turns (left vertical axis), and

difference with theoretical
0.000001000
expectation as per Eq. 3.17 0.05

(right vertical axis). The plot

Ynum. - Ytheor [cm]

shows two sinusoidal curves:
Y [cm]

a segmented one, thicker, 0 0.000000100

from numerical integration,

and a thinner one,
superimposed, from Eq. 3.17 -0.05
0.000000010

-0.1
0.000000001
0 1 2 3 4 5 6 7 8 9 10
s /circumference

3.9 RF Phase at the Accelerating Gap

According to Sect. 3.2.3 (Fig. 3.13), the RF is taken about half-way of the accel-
erating range, namely, referring to Fig. 3.39, .Trev = 0.131 µs and . f rf = 1/Trev =
7.633 MHz.
An input data file for this simulation is given in Table 3.27.
In a similar way to the diagrams in Fig. 3.13, the resulting . B(R) curve is given in
Fig. 3.44, the resulting ./\W (φ) curve in Fig. 3.45.

More turns are performed by changing the arguments under REBELOTE in the
input data file (Table 3.27), from 42 to 75 in the present case. The resulting energy gain
3.4 Solutions of Exercises of This Chapter 121

Table 3.26 Simulation input data file: for this wave number scan, the INCLUDE segment is taken
from Table 3.21

of the proton as a function of RF phase is shown Fig. 3.46. A first graph in Fig. 3.47
shows the evolution of its relative rigidity, namely D-1 as a function of distance, with
. D = Bρ(s)/B O R O and BORO = 64.624444 kG cm the reference rigidity as defined
under OBJET; a second graph shows its orbital radius as a function of distance.

3.10 The Cyclotron Equation

Cyclotron model settings of Exercise 3.3 are considered in questions (a) to (c), first:
two dees, double accelerating gap, uniform field . B = 0.5 T. The analytical field
modeling DIPOLE [16, lookup INDEX] is used.

(a) Simulation data file.

Acceleration is over the energy range .[0.2, 20] MeV, the maximum of
.cos(φ) (Fig. 3.14) is placed at . E k = E k,m = 10 MeV.
The cyclotron equation (Eq. 3.26) can be written under the form
[ ]
π ωr f
. cos φ = cos φi − (E 2 − E i2 ) − (E − E i ) (3.41)
q V̂ 2Mωrev
122 3 Classical Cyclotron

Tune dependence on energy, from zgoubi.MATRIX.out

1.02 0.5
νx νy (νx2+νy2)1/2

1 0.45

0.98 0.4
νx, (νx +νy2)1/2

0.96 0.35

νy
2

0.94 0.3

0.92 0.25

0.9 0.2

0.88 0.15
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5
kin. E [MeV]

Fig. 3.43 A scan of the energy dependence of the horizontal and vertical wave numbers. Markers
are from raytracing, solid lines are from theory (Eq. 3.18). The figure also shows that the raytracing
yields .ν 2R + ν y2 = 1, ∀E, as expected

where the index i denotes injection parameters, .φ is the phase of the RF at particle
arrival at the accelerating gap, .V̂ is the peak gap voltage, . E = E k + M is the total
energy with . M the rest mass. The value of . E k at the maximum of .cos φ is drawn
from .d(cos φ)/d E k = 0, namely
( )
ωrev
. E k,m = −1 M (3.42)
ωr f

Taking . E k,m = 10 MeV one gets

ωrev
. ≈ 1 + 0.106578, fr f ≈ 0.989454ωrev /2π = 7.542209 MHz
ωr f

The corresponding input data file is given in Table 3.28. Figure 3.48 shows the
case of two particles accelerated at a rate of 400 kV per turn, one resulting from an
initial phase at the gap of .φi = π/2 and reaches 20 MeV in about 52 turns, the other
resulting from an initial phase .φi = 3π/4 and reaches 20 MeV in about 64 turns. In
the latter case, the .π/4 phase shift results from an initial path length offset

δs = βcTrf /4 = 10.26647 cm
.
3.4 Solutions of Exercises of This Chapter 123

Table 3.27 Simulation input data file: accelerating a proton to get the evolution of RF phase The
[#S_60degSectorWIdx:#E_60degSectorWIdx] segment of Table 3.18 is INCLUDEd, here

Fig. 3.44 Radial Zgoubi|Zpop BZ (T) vs. Y (m)

dependence of the magnetic
0.5
field over the acceleration
range. The field is 0.5 T at a
reference radius . R0 = 0.5 m, 0.4
the slope results from the
index .k = −0.03. A graph 0.3
obtained using zpop: menu
7; 1/1 to open zgoubi.plt;
2/[2,32] for . B Z versus .Y ; 0.2
7 to plot
0.1

0.0 0.2 0.4 0.6 0.8 1.

as specified under OBJET (.βc = 0.020648c is the proton velocity at . E i = 200 keV),
yielding .δφ = ωrf δs / βc = π/4. A third curve in the figure is for to 200 kV voltage
and initial phase at gap .φ = π/2, in that case .cos(φ) reaches the value of 1 at about
4 MeV, 32 turns, and the particle starts decelerating.
124 3 Classical Cyclotron

Fig. 3.45 Span in phase of 0.12

the energy gain
./\W = q V̂ sin φ over the 0.1
acceleration range 200 keV 0.08
to 5 MeV. The vertical
separation of the two 0.06
./\W (φ) branches on the left
0.04

ΔW [MeV]
(./\φ < 0 above and ./\φ > 0
underneath) is artificial (a 0.02
“–($6–50)/10000.” “trick” in
the gnuplot script of 0
Table 3.27), for the sake of -0.02
clarity—they actually
superimpose -0.04
-0.06
-0.08
0.5 1 1.5 2 2.5 3 3.5 4
RF phase [rad]

Fig. 3.46 Span in phase of 0.15

the energy gain
./\W = q V̂ sin φ over an
acceleration and deceleration 0.1
cycle, starting from 200 keV.
The vertical separation of 0.05
./\W (φ) branches at the left
ΔW [MeV]

and right ends is artificial

0

-0.05

-0.1

-0.15
0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5
RF phase [rad]

(b) Energy-phase relationship.

A graph of the energy-phase relationship obtained by ray tracing, for .φi = 3π
4
and . π2
at the three different gap voltages .V̂ = 100, 200 and .400 kV, is given in Fig. 3.49,
together with theoretical expectations (Eq. 3.26).
3.4 Solutions of Exercises of This Chapter 125

Zgoubi|Zpop 1-D vs. s (m) Zgoubi|Zpop Y (m) vs. s (m)

4.5
0.7
4.

3.5 0.6
3.
0.5
2.5

2. 0.4

1.5
0.3
1.

0.5 0.2

0 50 100 150 200 0 50 100 150 200

Fig. 3.47 Left: relative rigidity offset of the proton as a function of distance around the ring,
accelerating over half the path, and subsequently decelerating back to the initial energy, under the
effect of the cumulated phase-shift. Right: increase first and decrease next of the orbital radius as a
function of azimuthal distance

3.11 Cyclotron Extraction

(a) Distance between turns.

Simulation input data of Exercise 3.3, Table 3.8, can be referred to as a guidance to
build the present simulation file.
A proton is accelerated in 26 turns, in a uniform field . B0 = 0.5 T, from
20 keV (rigidity . B O R O × D = 0.064624444 × 0.3162126 = 0.0204350634608
T m, injection radius .Y0 = B R/B0 = 4.08701269216 cm) to 5.02 MeV. The RF
phase is ignored thus CAVITE[IOPT = 3] is used, with a 100 kV gap voltage. The
input data file for this simulation is given in Table 3.29.
The accelerated orbit and the distance./\R between turns are displayed in Fig. 3.50.
Theoretical expectation (Eq. 3.27) in the case of slow acceleration (typically, the fixed
energy closed orbit configuration of Fig. 3.21) is also displayed, for comparison.

(b) Beam losses.

Indications to solve this exercise:
– a beam with Gaussian momentum distribution and rms momentum spread .δp/ p =
10−3 can be defined using MCOBJET,
– use REBELOTE to accelerate over a given number of turns,
– an extraction septum placed half-way between two successive turns can be simu-
lated using COLLIMA, placed after REBELOTE (the execution pointer will quietly
continue beyond REBELOTE do-loop once the latter is completed). COLLIMA
counts particles stopped. FAISTORE (or FAISCNL) can be placed after COLLIMA,
to log particle data: particles stopped by COLLIMA have their IEX tag set to IEX =
–4 [16, lookup COLLIMA].
Change the value of NPASS under REBELOTE for a different number of accel-
erated turns, and COLLIMA positioning data accordingly.
126 3 Classical Cyclotron

Table 3.28 Simulation input data file: the cyclotron equation (Eq. 3.26). This requires a uni-
form field, for that the [#S_60degSectorUnifB:#E_60degSectorUnifB] segment of Table 3.6 is
INCLUDEd, here. Note the PRINT instruction under CAVITE: it causes a print out of CAVITE
computational data in zgoubi.CAVITE.out, during the ray tracing, including RF phase and ion
energy which can then be plotted (gnuplot script below, called by SYSTEM, and Fig. 3.48). The
second particle under OBJET is launched on the closed orbit, its initial phase at the voltage gap
is .π/2. The first and third particles leave with an initial longitudinal shift .δs = ∓10.26647 cm at
OBJET resulting in .π/4 and .3π/4 initial phase at the voltage gap

(c) Change the field index.

The cyclotron model of Table 3.21 is used here, reference field . B0 = 5 kG on the
200 keV orbit, and field index k = –0.03. A proton is accelerated over 26 turns, from
20 keV to 5.02 MeV, as in question (a). The 20 keV closed orbit radius (taken as the
injection radius) differs from question (a) due to the index k = –0.03, and can be
found using a FIT procedure (Table 3.30); it comes out to be .Y0 = 4.0040586 cm.
The input data file for this exercise is given in Table 3.31.
The resulting proton trajectory is displayed in Fig. 3.51 (the gnuplot script given in
Table 3.31 is used). The greatly different accelerated orbit in this case, compared to the
uniform field case in (a) (Fig. 3.50), results from a modulation of the distance between
turns, which is an effect of the oscillation motion undergone by the accelerated orbit
3.4 Solutions of Exercises of This Chapter 127

Zgoubi|Zpop Kinetic Energy (MeV) vs. Turn

20

δ s = 10.266 δ s =0
15 400 kV 400 kV

10

δ s =0
5 200 kV

10 20 30 40 50 60 70

Fig. 3.48 Proton energy versus turn, case of (Table 3.28) voltage 400 kV/turn, two protons with
initial phase respectively .π/2 (.δs = 0) and .3π/4 (.δs = 10.26647 cm), which make it up to 20 MeV
and beyond. The third case, voltage 200 kV/turn, initial phase .π/2 (.δs = 0), features a maximum
energy of 4 MeV and deceleration from there on. A graph obtained using zpop: menu 7; 1/5 to read
from zgoubi.fai; 2/[39,2] for .Y versus turn

Fig. 3.49 A graph of the

 /2, 100kV
energy dependence of the 1.5 200kV
arrival phase at the voltage 400kV
gap, for a few different  /4, 100kV
values of gap voltage .V̂ and 1 200kV
initial phase .φi . Markers are 400kV
from raytracing, using the
0.5
input data file of Table 3.28
cos()

repeatedly for the various

values of .V̂ and .φi . 0
Superimposed solid lines are
from theory (Eq. 3.26 and
Fig. 3.14) -0.5

-1
0 5 10 15 20
Ek [MeV]

(around the local on-momentum half-circle orbit arc). This effect may be exploited
to increase extraction efficiency, by causing such a radial modulation as to maximize
turn separation at the location of the septum [17].

(d) Optimize extraction.

The modulation is minimized (or enhanced possibly, at the last turn, for minimized
losses at extraction) by optimizing the injection conditions .(x0 , x0, ).
128 3 Classical Cyclotron

Table 3.29 Simulation input data file: accelerating a proton to check evolution of./\R/R, in a dipole
field with index. The #S_180degSectorUnifB to #E_180degSectorUnifB segment of Table 3.6 is
INCLUDEd

3.12 Acceleration and Extraction of a 6-D Polarized Bunch

This simulation can be set up using material drawn from previous exercises. It is not
fully developed here, guidelines are given.
3.4 Solutions of Exercises of This Chapter 129

0.8
10

0.6

0.4

0.2
YLab [m]

ΔR [cm]
0 num.
0 0.2 0.4 0.6 0.8 theory

-0.2

-0.4

-0.6

-0.8 1
-0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 10 20 30 40 50 60
XLab [m] R [cm]

Fig. 3.50 Left: accelerated orbit from 20 keV to 5.02 MeV, at a rate of 200 keV per turn over 26
turns, in a uniform field. The thick horizontal line (colored) figures the accelerating gap. Right: the
resulting dependence of orbit separation ./\R on radius, from raytracing (markers) and from theory
(solid line); the theoretical curve assumes small dE (adiabatic acceleration, concentric orbits), which
is not quite the case here with ./\E = 200 keV/turn

Table 3.30 Simulation input data file: finding the 20 keV injection radius in the presence of a
non-zero index k, using FIT The INCLUDE segment is taken from Table 3.21

The cyclotron simulation hypotheses of Exercise 3.10a are considered, the input
data file for this exercise can be built from that of Table 3.28, with a few modifications,
namely:
– downstream of REBELOTE, add a 1 meter DRIFT: an embryo of an “high energy
line” into which the bunch is steered at extraction;
– that DRIFT is preceded by CHANGREF to center the current reference frame on
the final coordinates .Y and .T of the accelerated orbit; the latter have to be determined
by prior raytracing;
– add histograms (to be logged in zgoubi.res) for observation of transverse and
longitudinal particle coordinate densities in the bunch at extraction. This uses HISTO,
as many times as needed.
130 3 Classical Cyclotron

Table 3.31 Simulation input data file: accelerating a proton to check evolution of ./\R/R, in a
dipole field with index. The [#S_60DegSectorR200:#E_60DegSectorR200] segment of Table 3.21
is INCLUDEd
References 131

1 10

0.8

0.6

0.4
1

0.2
YLab [m]

ΔR [cm]
0
0 0.2 0.4 0.6 0.8 1

-0.2
0.1
-0.4

-0.6

-0.8

-1 0.01
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 10 20 30 40 50 60 70 80
XLab [m] R [cm]

Fig. 3.51 Left: accelerated orbit from 20 keV to 5.02 MeV, at a rate of 200 keV per turn over 26
turns, in a dipole field with index. The thick horizontal line (colored) figures the accelerating gap.
Right: the resulting dependence of orbit separation ./\R on radius, observed at the second gap

References

1. A. Sessler, E. Wilson, Engines of Discovery. A Century of Particle Accelerators (World Scien-

tific, 2007)
2. E.O. Lawrence, M.S. Livingston, Phys. Rev. 37, 1707 (1931); Phys. Rev. 38, 136, (1931); Phys.
Rev. 40, 19 (1932)
3. Credit: Lawrence Berkeley National Laboratory. The Regents of the University of California,
Lawrence Berkeley National Laboratory
4. E.O. Lawrence, M.S. Livingston, The production of high speed light ions without the use of
high voltages. Phys. Rev. 40, 19–35 (1932)
5. M.S. Livingston, E.M. McMillan, History of the cyclotron. Phys. Today 12(10) (1959). https://
escholarship.org/uc/item/29c6p35w
6. H.E. Bethe, M.E. Rose, Maximum energy obtainable from cyclotron. Phys. Rev. 52, 1254
(1937)
7. F.T. Cole, O Camelot ! A Memoir of the MURA Years (1994). https://accelconf.web.cern.ch/
c01/cyc2001/extra/Cole.pdf
8. L.H. Thomas, the paths of ions in the cyclotron. Phys. Rev. 54, 580, (1938); M.K. Craddock, AG
focusing in the thomas cyclotron of 1938, in Proceedings of PAC09 (Vancouver, BC, Canada,
FR5REP1)
9. T. Stammbach, Introduction to cyclotrons. CERN accelerator school, cyclotrons, linacs and
their applications. IBM International Education Centre, La Hulpe, Belgium, 28 April-5 May
1994. Fig. 8; T. Stammbach, Introduction to Cyclotrons. CERN Yellow Report 96-02 (1996),
Figure 8, page 15. Copyright/License CERN CC-BY-3.0 https://creativecommons.org/licenses/
by/3.0, no change to the material
10. E. Baron et al., The GANIL injector, in Proceedings of the 7th International Conference on
Cyclotrons and their Applications (Zürich, Switzerland, 1975). http://accelconf.web.cern.ch/
c75/papers/b-05.pdf
11. L.B. Cohen, Cyclotrons and synchrocyclotrons, in Encyclopedia of Physics, Vol. XLIV, Nuclear
Instrumentation I, ed. by S. Flūgge (Springer, 1959)
12. J. Le Duff, Longitudinal Beam Dynamics in Circular Accelerators (CERN Accelerator School,
Jyvaskyla, Finland, 1992)
13. B.W. Montague, Polarized beams in high energy storage rings. Phys. Rep. (Rev. Sect. Phys.
Lett.) 113(1), 1–96 (1984)
132 3 Classical Cyclotron

14. T. Roser, A. Zelensky, Private communication, BNL, June 2021. Günther Clausnitzer: History
of Polarized Ion Source Developments, in International Workshop on Polarized Ion Sources and
Polarized Gas Jets (KEK, Tsukuba, Japan, 1990). KEK Report 90–15, November 1990, ed. by
Y. MORI. https://inis.iaea.org/collection/NCLCollectionStore/_Public/22/051/22051667.pdf
15. F. Méot, Spin dynamics, in Polarized Beam Dynamics and Instrumentation in Particle Acceler-
ators, USPAS Summer 2021 Spin Class Lectures (Springer Nature, Open Access, 2023). https://
link.springer.com/book/10.1007/978-3-031-16715-7
16. F. Méot, Zgoubi Users’ Guide. https://www.osti.gov/biblio/1062013-zgoubi-users-guide.
Sourceforge latest version: https://sourceforge.net/p/zgoubi/code/HEAD/tree/trunk/guide/
Zgoubi.pdf
17. T. Stammbach, Introduction to Cyclotrons, in CERN Accelerator School, Cyclotrons, Linacs
and Their Applications (IBM International Education Centre, La Hulpe, Belgium, 1994)

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0
International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing,
adaptation, distribution and reproduction in any medium or format, as long as you give appropriate
credit to the original author(s) and the source, provide a link to the Creative Commons license and
indicate if changes were made.
The images or other third party material in this chapter are included in the chapter’s Creative
Commons license, unless indicated otherwise in a credit line to the material. If material is not
included in the chapter’s Creative Commons license and your intended use is not permitted by
statutory regulation or exceeds the permitted use, you will need to obtain permission directly from
the copyright holder.
Chapter 4
Relativistic Cyclotron

Abstract This chapter introduces the AVF (azimuthally varying field), isochronous,
relativistic cyclotron, and to the theoretical material needed for the simulation exer-
cises. A brief reminder of the historical context is followed by further basic theoretical
considerations leaning on the cyclotron concepts introduced in Chap. 3 and including
– Thomas focusing and the AVF cyclotron,
– positive focusing index,
– isochronous optics,
– separated sector cyclotrons,
– spin dynamics in an AVF cyclotron.

Simulation exercises use optical elements and keywords met earlier: the analytical
field modeling DIPOLE, TOSCA in case using a field map is preferred, CAVITE to
accelerate, SPNTRK to solve spin motion, FAISCEAU, FAISTORE, FIT, etc. The
exercises further develop on radial and spiral sector magnets, edge focusing and
flutter, isochronous optics, separated sector ring cyclotrons, and their modeling in
DIPOLE, DIPOLES and other CYCLOTRON keyword capabilities.

Notations Used in the Text

.B; .B0 magnetic field; at a reference radius .R0

.B; .BR ; .Bθ ; .By field vector; radial, azimuthal and axial components
.BR = p/q; BR0 magnetic rigidity; reference rigidity
.C; .C0 closed orbit length, .C = 2π R; reference, .C0 = 2π R0
.E ion energy, .E = γ m0 c2
EFB effective Field Boundary
( )1/2
.F ; .F azimuthal field form factor; flutter, .F = <(F <F
−<F >)2 >
> 2

.frev , .frf revolution and RF voltage frequencies

.h harmonic number, an integer, .h = frf /frev
.k =
R dB
B dR
geometric index, a global quantity
.m; .m0 ; M ion mass; rest mass; in units of MeV/c.2

© The Author(s) 2024 133

F. Méot, Understanding the Physics of Particle Accelerators, Particle Acceleration
and Detection, https://doi.org/10.1007/978-3-031-59979-8_4
134 4 Relativistic Cyclotron

.n = Bρ ddBρ focusing index, a local quantity

.p; p0 ion momentum vector; reference momentum
.q ion charge
R; .R0 ; .RE average radius of equilibrium orbit; .R = C/2π ; .R(p = p0 ); .R(E)
.R radial field form factor
.RF Radio-Frequency
.s path variable
.Trev , .Trf revolution and accelerating voltage periods
.v ion velocity
V(t); .V̂ oscillating voltage; its peak value
[ ]
(∗)
x' , y' radial and axial coordinates . (∗)' = dds
.α trajectory deviation, or momentum compaction
.β = v/c; .β0 ; .βs normalized ion velocity; reference; synchronous
.γ = E/m0 c
2
Lorentz relativistic factor
.Δp, .δp momentum offset
.ε wedge angle
.ϵR strength of a depolarizing resonance
.εu Courant-Snyder invariant (.u : x, y, l), ...)
.ζ spiral angle of a spiral sector dipole EFB
.θ azimuthal angle
.φ; φs phase of oscillating voltage; synchronous phase

4.1 Introduction

Isochronous cyclotrons are in operation today by the thousands, tens are produced
each year. Applications include production of radio-isotopes mostly, proton therapy
(Fig. 4.1), high power beams for accelerator-driven systems, secondary particle beam
production (Fig. 4.2), and more [1]. The technology and its applications are fostered
by cryogeny and high fields which further allow compactness (Fig. 4.1) as well as
highest beam rigidities (Fig. 4.3).
At the origin of the evolution of the cyclotron technology, which led to the AVF
innovation in the late 1930s, is the energy limitation of the classical cyclotron, at
a few tens of MeV/nucleon (Chap. 3). Axial focusing in the latter results from the
slow decrease of the guiding field with radius in the wide gap between the elec-
tromagnet poles. That negative field index .−1 < k < 0 (Eqs. 3.11, 3.12) results in
both radial and axial periodic stability (Eq. 3.18). Isochronism requires instead the
field to increase with radius, i.e. a field index .k > 0, a consequence of .B(R) ∝ γ (R)
(Sect. 4.2). The AVF concept by L.H. Thomas in 19381 [5] (Fig. 4.4), solved the prob-
lem: AVF entails axial periodic stability as long as the field modulation parameter

1The very L.H. Thomas of the Thomas-BMT spin motion differential equation, author of the eight
years earlier Nature article [7].
4.1 Introduction 135

Fig. 4.1 COMET protontherapy cyclotron at PSI. A 250 MeV, 500 nA, 4-sector isochronous AVF
cyclotron. The spiral poles enhance axial focusing. A 3 m diameter superconducting coil provides
the dipole field [2]

F > βγ (Sect. 4.2.1). The vertical defocusing effect which the radially increasing
.
field causes is compensated by the focusing effect of the AVF. Spiral pole geometry
was further introduced in 1954 [8] to increase axial focusing, so allowing greater .k
and isochronous acceleration to higher energy (Sect. 4.2.2). It took some time, until
the late 1950s (see Stammbach’s Fig. 3.4), for Thomas’ concept to make its way
136 4 Relativistic Cyclotron

Fig. 4.2 PSI 590 MeV ring cyclotron delivers a 1.4 MW proton beam. Acceleration takes .∼180
turns; extraction efficiency is .>99.99%; overall diameter is 15 m. Beam is used for the production
of secondary neutron and muon beams [3]

and lead up to practical realisations2 [9, 10]. AVF cyclotrons were constructed to
accelerate all sorts of ions whereas classical cyclotrons tended to leave the scene
(Fig. 3.4). Applications included material science, radiobiology, production of sec-
ondary beams, and more. Polarized ion beams became part of the landscape as well
from the moment polarized ion sources were made available [11].
The separated sector method was developed in the early 1960s, instances are
today’s high power PSI 590 MeV spiral sector cyclotron (Fig. 4.2), brought into oper-
ation in 1974, and its injector-II, a radial-sector design (Fig. 4.5). Iron-free regions
between separated sector dipoles allows room for multiple high-Q RF resonators
thus greater turn separation at extraction, for higher efficiency extraction systems and

2 One can read for instance, in 1959s Ref. [10], Cyclotrons and Synchrocyclotrons, regarding engi-
neering aspects, “Also, no consideration is given to the AVF cyclotron, since none of this type has
reached the advanced design stage”.
4.1 Introduction 137

Fig. 4.3 RIKEN K2500, superconducting coil, separated-sector, 8,300 ton ring cyclotron [4]. The
dipole field is 3.8 T, rigidity 8 T m, diameter 18.4 m. Beam injection radius is 3.56 m, extraction
radius is 5.36 m. The cyclotron is part of a radioactive ion beam accelerator complex

0.8
10
9 0.6
8
7
10 6 0.4
9 5
8
7 4
6 3 0.2
BZ [kG] 5
4 2
Ylab [m]

3 1
2 0
0 0 0.2 0.4 0.6 0.8
1
0
-0.8
-0.2
-0.6
-0.4
-0.8 -0.2 -0.4
-0.6
-0.4 0
-0.2 0.2 X [m] -0.6
0
0.2 0.4
Y [m] 0.4 0.6
0.6 -0.8
0.8 0.8 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8
Xlab [m]

Fig. 4.4 A 4-periodic AVF cyclotron design (after Ref. [5]). Left: mid-plane azimuthally modulated
field. Right: closed orbits around the cyclotron feature azimuthally varying curvature, greater on
the hills, weaker in the field valleys
138 4 Relativistic Cyclotron

Fig. 4.5 PSI injector II, four separated radial sectors, 0.87 MeV injection energy, accelerates protons
to 72 MeV in about 100 turns [6]. The drifts include the 50.7 MHz accelerating RF system and a
flattop cavity. Injection is from the top, in the central region

thus higher beam current, and for the insertion of beam instrumentation. Cyclotron
energy subsequently increased, up to the present days near-GeV range. Cryogeny was
introduced in the early 1960s at the Michigan State University superconducting coil
K500 cyclotron3 [12]. Superconducting technology allows higher field and reduction
of size, culminating today with RIKEN’s K2500 SRC (Fig. 4.3 and Table 4.1).

4.2 Basic Concepts and Formulæ

Mass increase with energy causes loss of synchronism in the classical cyclotron,
and the required negative field index (decreasing guiding field with radius) for axial
periodic motion stability adds to the effect. Isochronism instead, i.e., constant .ωrev =
qB/γ m0 , given orbit radius .R = βc/ωrev , leads to positive index

R ∂B β ∂γ
.k= = = β2γ 2 (4.1)
B ∂R γ ∂β

3 .K= E A/Z 2 , with A the number of mass, Z the number of charge, is a measure of the equivalent
proton energy, 500 MeV in this case.
4.2 Basic Concepts and Formulæ 139

Table 4.1 A comparison between an AVF and a separated sector cyclotron of same energy, 72 MeV,
namely, the former AVF injector and the present Injector II of PSI high power cyclotron, after
Ref. [11, p. 126]
AVF Separated sector
Injection energy keV 14 870
Extraction energy MeV 72 72
Beam current mA 0.2 1.6
Magnet single dipole 4 sectors
Weight ton 470 .4 × 180

Dipole gap height mm 240–450 35

.⟨B⟩; .Bmax T 1.6; 2 0.36; 1.1
RF system ◦
180. dees 2 resonators
Accelerating voltage kV .2 × 70 .4 × 250

RF MHz 50 50
Normalized beam .π µm 2.4; 1.2 1.2; 1.2
emittance, hor.; vert.
Beam phase width deg 16–40 12
Energy spread % 0.3 0.2
Turn separation at mm 3 18
extraction

requiring .k to follow the energy increase: the weak focusing condition .−1 < k < 0
can not be satisfied, transverse periodic stability is lost.
Isochronism requires the revolution period .Trev = 2π γ m0 /qB to be momentum
independent; under this condition, differentiating this expression yields the radial
field dependence
B0
.B(R) = γ (R) (4.2)
γ0

with .B0 and .γ0 some reference conditions,

This led H.A. Bethe and M.E. Rose to conclude, in 1938, “... it seems useless
to build cyclotrons of larger proportions than the existing ones... an accelerating
chamber of 37 cm radius will suffice to produce deuterons of 11 MeV energy which
is the highest possible...” [13]. And F.T. Cole to comment, “If you went to graduate
school in the 1940s, this inequality [.−1 < r(dB/dr)/B < 0] was the end of the
discussion of accelerator theory” [14, Sect. 1.4].

4.2.1 Thomas Focusing

Whereas the classical cyclotron approach assumed revolution symmetry of the field,
a 1938 publication stated: “[...] a variation of the magnetic field with angle, [...] of
140 4 Relativistic Cyclotron

Fig. 4.6 Pole shaping in an AVF cyclotron, an electron model, here [15]. The focusing pattern is
FfFfFf, an alternation of strong (hill regions) and weak (valleys) radial focusing [16]

order of magnitude v/c; together with nearly the radial increase of relative amount
.
2
v /c of Bethe and Rose; gives stable orbits that are in resonance and not defo-
1 2 2

cused” [5]. In other words, AVF in proper amount (Fig. 4.4) compensates the axial
defocusing resulting from the increase of the field with radius (Eq. 4.2). Azimuthal
field modulation and radial increase may be obtained by shaping the magnet poles,
as illustrated in Fig. 4.6.

Azimuthal Field Modulation, Flutter

A simple approach to the .2π/N -periodic axial symmetry and field modulation may
assume a sinusoidal azimuthal form factor

.F (θ ) = 1 + f sin(N θ ) (4.3)

This is the case in Fig. 4.4, for instance. The mid-plane field can thus be expressed
under the form
.B(R, θ ) = B0 R(R) F (θ ) (4.4)
4.2 Basic Concepts and Formulæ 141

with .R(R) the radial dependence. The orbit curvature varies along the . 2π N
-periodic
ρ(s) dB
orbit, this requires distinguishing between the local focusing index.n = B(s) d ρ and the
geometrical index .k (Eq. 4.1), a global quantity which determines the wave numbers
(Eq. 4.6). A “flutter” factor can be introduced to quantify the effect of the azimuthal
modulation of the field on the focusing,

hard
( )1/2 ( )1/2
< (F − < F >)2 > ed ge R
F=
. −→ −1 (4.5)
< F >2 ρ
∮
where .< ∗ >= (∗) d θ/2π . If the scalloping of the orbit (i.e., its excursion in the
vicinity of .R) is of small amplitude, then .R ≈ ρ and, accounting for the isochronism
condition (Eq. 4.1), approximate values of the wave numbers write
√ isochr.
√ isochr.
√
ν ≈
. R 1 + k = γ, νy ≈ −k + F 2 = −β 2 γ 2 + F 2 (4.6)

Thus the horizontal wave number increases during acceleration, linearly with energy,
whereas in the absence of countermeasure the axial wave number would decrease
- see Sect. 4.2.2. An additional property is

hard
ed ge R
ν 2 + νy2 = 1 + F 2
. R −→ (4.7)
ρ

The flutter allows designing .−k + F 2 > 0 (whereas .k > 0), so ensuring peri-
odic stability of the axial motion. In √
the hypothesis of a sinusoidal azimuthal field
modulation (Eq. 4.3) one has .F = f / 2 and
√
ν ≈
. y −k + f 2 /2, νR2 + νy2 = 1 + f 2 /2 (4.8)

Off-Momentum Orbit

The dispersion function .D = δx/ δp/p in the revolution symmetry field (Eq. 3.20)
has the form .D = 1+k
R0
. Given the isochronous condition .k = β 2 γ 2 it can be written

R
D=
. (4.9)
γ2

An alternate approach consists in considering that, with the isochronism condition

2π R/βc = Trev , a constant, one gets . dR
.
R
= dββ = γ12 dp
p
.
AVF Modeling
A numerical approach to the azimuthal modulation beyond the simple sine modula-
tion of Eq. 4.3, is discussed in Sect. 14.3.3 (Eqs. 14.11, 14.15). It provides a modeling
of .F (θ ) over the whole beam excursion area, possibly including an R-dependence,
142 4 Relativistic Cyclotron

0
ε>

ε>
0
field is 120 deg

t
missing

rbi
do
se
clo O
field is
added 60 deg

Fig. 4.7 A 120 deg bending of the closed orbit (curvature center at O) is ensured by a 60 deg
bending sector. This results in a wedge angle (.ε > 0 by convention in this configuration) in the
transition regions between valleys and hills, which causes a decrease of the radial focusing (solid
incoming trajectories, compared to dotted ones), and axial focusing under the effect of the trajectory
angle to the azimuthal field component

F (R, θ ). The method ensures the continuity of .F (R, θ ) and its derivatives, between
.
neighboring magnetic sectors. It is resorted to in the simulation exercises.
Wedge Focusing
In the entrance and exit regions of a bending sector, closed orbits are at an angle to the
iso-field lines, this causes “wedge focusing”, an effect sketched in Fig. 4.7: with posi-
tive wedge angle .ε, case of the AVF configuration, radial focusing decreases whereas
the angle of off mid-plane particle velocity vector to the azimuthal component of the
field in the wedge region causes axial focusing.

4.2.2 Spiral Sector

Spiral sector geometry was introduced in 1954 in the context of fixed field alternating
gradient accelerator (FFAG) studies [8], and found application in cyclotrons (as in
PSI’s COMET cyclotron, Fig. 4.1). Spiraling the edges (Fig. 4.8) results in stronger
axial focusing (Eq. 4.12) compared to a radial sector (Eq. 4.6), it also permits an
increase of the wedge angle with radius, so maintaining proper compensation of an
increase of .k(R) (Eq. 4.1). In a spiral sector bend the wedge angle is positive on one
side of the sector, negative on the other side (Fig. 4.8), with a global axial focusing
resultant. In a similar approach to the periodic field modulation in a radial sector
(Eq. 4.3), a convenient approach to the spiral sector AVF uses azimuthal form factor
[ ( )]
R
.F (R, θ ) = 1 + f sin N θ − tan(ζ (R)) ln (4.10)
R0
4.2 Basic Concepts and Formulæ 143

Fig. 4.8 Geometrical

parameters of a spiral sector
dipole. The center of the ring
is at O, .ζ is the spiral angle
(increasing with radius), .ε is
ζ
the wedge angle. In the hard ε
edge field model, a line of
constant field inside the
sector is an arc of radius .R;
thus the curvature radius .ρ ρ
varies along the closed orbit π/N
in the dipole
R

with the spiral angle .ζ (R) an increasing function of radius .R, whereas the mid-plane
field now writes under the form

B(R, θ ) = B0 R(R) F (R, θ )

. (4.11)

The local magnet edge geometry at R satisfies .r = r0 exp(θ/ tan(ζ )), a logarithmic
spiral centered at the center of the ring, with .ζ the angle between the tangent to the
spiral edge and the ring radius (Fig. 4.8). This results in a larger contribution of the
flutter term in the axial wave number,
√
.νy ≈ −k + F 2 (1 + 2 tan2 ζ ) (4.12)

As the field index k increases with .R to ensure isochronism (Eq. 4.1), the spiral angle
follows so to maintain.−k + F 2 (1 + 2 tan2 ζ ) > 0. A limitation here is the maximum
spiral angle achievable, obviously .ζ → 90 deg.
As an illustration, in TRIUMF cyclotron .ζ reaches 72 deg in the 500 MeV region
(from zero in the 100 MeV region) whereas .1 + 2 tan2 ζ increases to 20 (from 1 in
the 100 MeV region) and compensates a low .F < 0.07 (down from .F = 0.3). In PSI
590 MeV cyclotron .ζ reaches .35o on the outer radius. Most isochronous cyclotrons
of a few tens of MeV use spiral sectors to benefit from the more efficient axial
focusing [16].
More can be found in the scaling FFAG chapter (Sect. 10.2.2) regarding the spiral
sector, and regarding its numerical simulation.
144 4 Relativistic Cyclotron

4.2.3 Isochronism

In the hypothesis of isochronism, the revolution angular frequency satisfies .ωrev =

cβ(γ )/R(γ ) = constant. An orbital radius .R∞ = c/ωrev is reached asymptotically
as .β = v/c = R/R∞ → 1. In terms of the RF and harmonic number,
c
R∞ = h
. (4.13)
ωrf
( )−1/2
Given .BR∞ = γ m0 c/q and using .γ = 1 − (R/R∞ )2 , the radial dependence of
the field can be expressed in terms of .R∞ , namely,

B0 m0 ωrev m0 ωrf
B0 R(R) = γ B0 = √
. with B0 = = (4.14)
1 − (R/R∞ )2 q q h

and goes to infinity with .R → R∞ . For protons for instance, with

m0 /q = 1.6726 × 10−27 [kg] / 1.6021 × 10−19 [C] ≈ 10−8 ,.BR∞ [T m] = γ m0 c/q ≈
.
3γ . A typical value for .R∞ can be obtained assuming for instance an upper .γ = 1.64
(600 MeV) in a region of upper field value .B = 1.64 T, yielding .R∞ ≈ 3 m.

Radial Field

From Eq, 4.14 it results that the radial field form factor of Eqs. 4.4, 4.11 can be
written ( ( ) )−1/2
R 2
.R(R) = 1− (4.15)
R∞

A possible approach consists in using the Taylor expansion of .R(R) (within the limits
of radius of convergence of that series), namely
( )2 ( )4 ( )6
1 R 3 R 5 R
.R(R) = 1 + + + + ... (4.16)
2 R∞ 8 R∞ 16 R∞

The coefficients in this polynomial in .R/R∞ are the field index and its derivatives,
they can be a starting point for further refinement of the isochronism, including for
instance side effects of the azimuthal field form factor .F (R, θ ) (Eqs. 4.3, 4.10).
The radial field index.k(R) in the AVF cyclotron is designed to satisfy the condition
of isochronism (Eq. 4.1). However, reducing the RF phase slip over the acceleration
cycle substantially below .±π/2 requires a tolerance below .10−5 on field value over
the orbit excursion area. This tight constraint requires pole machining, shimming,
and other correction coil strategies in order to satisfy Eq. 4.1.
4.2 Basic Concepts and Formulæ 145

Fast Acceleration

Fixed field and fixed RF allow fast acceleration, the main limitation is in the amount
of voltage which can be implemented around the ring. The voltage per turn reaches
4 MV for instance at the PSI 590 MeV ring cyclotron, where bunches are accelerated
from 72 MeV to 590 MeV in less than 200 turns.
Harmful resonances may have to be crossed as wave numbers vary during acceler-
ation, including the “Walkinshaw resonance” .νR = 2νy as .νR ≈ γ whereas the axial
wave number spans .νy ≈ 1–∼1.5. This coupling resonance may result in an increase
of vertical beam size and subsequent particle losses, fast crossing mitigates the effect.
Fast acceleration improves extraction efficiency, as the turn separation .dR/dn is
proportional to the energy gain per turn (Sect. 4.2.4).

4.2.4 Cyclotron Extraction

The minimum radial distance between the last two turns, where the extraction septum
is located, is imposed by beam loss tolerances, which in some cases (high power
beams for instance) may be tight, in the .10−4 range or less. Space charge in particular
matters, as it increases the energy spread, and thus the radial extent of a bunch. In
the relativistic cyclotron the separation between two consecutive turns satisfies

γ ΔE R
ΔR ≈
. (4.17)
γ + 1 E νR2

with .ΔE the effective acceleration rate per turn. This indicates that greater turn
separation at extraction results from increased ring size. As a matter of fact, size is a
limitation to intensity in small cyclotrons. It also indicates that extraction efficiency
may be increased by moving the radial wave number closer to .νR = 1.

4.2.5 Resonant Spin Motion

In the quasi-uniform, quasi vertical field .B ≈ By of a classical cyclotron dipole, spins

quietly perform .Gγ precessions around a vector .ωsp || B (Eq. 3.30) as the particle
velocity completes a .2π precession around the ring (Sect. 3.2.5) [17].
More is liable to happen in the AVF cyclotron, due to the strong radial field index
(Eq. 4.1) and to the azimuthal field modulation (Eqs. 4.3, 4.10): the azimuthal and
radial field components .Bθ and .BR are non-zero out of the median plane, .B(R, θ, y)
may locally depart from vertical in a substantial manner, and so will the local pre-
cession vector .ωsp (R, θ, y). The latter varies periodically in addition, as the particle
undergoes periodic vertical motion about the median plane. Resonance between spin
precession (characterized by spin tune .νsp = Gγ , Eq. 3.33) and periodic perturbing
146 4 Relativistic Cyclotron

field components (characterized by the axial wave number .νy , Eqs. 4.6, 4.12) occurs
if the two motions feature coinciding frequencies. This condition can be expressed
under the form

ν ± νy = integer or, equivalently Gγ = integer ± νy

. sp (4.18)

The spin precession axis .ωsp moves away from the vertical as the spin motion gets
closer to resonance (during acceleration as .Gγ varies for instance), to end up in the
median plane on the resonance [18, Sect. 3.6].
Consider
⟨ ⟩ now an ion bunch, away from any depolarizing resonance. Its polariza-
tion is . Sy , the average of the projection of the spins on the vertical. If a depolarizing
resonance is crossed during acceleration, the initial polarization (far upstream of the
resonance; index i) and final polarization (far downstream of the resonance; index f)
satisfy the Froissart-Stora law [19],

⟨ ⟩ π |ϵR |2
Sy f −
.⟨ ⟩ = 2e 2 a − 1 (4.19)
Sy i

where .|ϵR | is the strength of the resonance: a measure of the strength of the depolar-
izing fields, its calculation is addressed in a next chapter; .a is the resonance crossing
speed,
dγ d νy
.a = G ± (4.20)
dθ dθ
The
⟨ ⟩ Froissart-Stora
⟨ ⟩ formula indicates that, if the resonance is crossed slowly (.a → 0),
. Sy f / Sy i → −1: spins quietly follow the flipping motion of the precession axis,
⟨ ⟩ ⟨ ⟩
polarization is flipped and preserved. If the crossing is fast (.a → ∞), . Sy f / Sy i →
0,
|⟨ polarization
⟩| is unaffected. Intermediate crossing speeds cause polarization loss:
.| Sy | ends up smaller after the resonance.

4.3 Exercises

Exercises 4.2–4.4 use a field map, designed in Exercise 4.1, to simulate an AVF
cyclotron dipole. Note that they can be performed using DIPOLE[S] analytical field
model instead, as in Exercise 4.5 (a similar simulation which can be referred to is
Exercise 3.2, Classical Cyclotron chapter). As a reminder, regarding the interest of
one or the other of the two methods: field maps allow close to real field models
(a measured field map for instance, or from a magnet computer code); using an
analytical field model allows more flexibility regarding magnet parameters, which
can for instance be optimized using a matching procedure.
4.3 Exercises 147

Note: some of the input data files for these simulations are available in zgoubi
sourceforge repository at
[pathTo]/branches/exemples/book/zgoubiMaterial/cyclotron_relativistic/

4.1 Modeling Thomas AVF Cyclotron

Solution: 4.1.
In this exercise a 2D mid-plane field map is built, inspired from Thomas’s 1938 arti-
cle [5]. The method to build the map is that of Exercise 3.1, TOSCA[MOD.MOD1=
22.1] keyword is used to raytrace through and derive the optical parameters of the
4-period AVF cyclotron.
(a) Construct a .360◦ 2D map of the median plane field .BZ (R, θ ), simulating the
field in the 4-period Thomas cyclotron of Fig. 4.4, assuming the following:

– .BZ (R, θ ) = B0 [ 1 + f sin(4(θ − θi )) ] (Eq. 4.3), with .θi some arbitrary origin of
the azimuthal angle, to be determined. Hint: depending on .θi value, the closed
orbit may be at an angle to the polar radius, as seen in Fig. 4.4; in that case
TOSCA[MOD.MOD1=22.1] would require non-zero in and out positioning angles
TE and TS, to be determined and stated using KPOS option [20]; instead, a proper
choice of .θi value allows a simpler TE .= TS .= 0;
– an average axial field .B0 = 0.5 T on the 200 keV radius (the latter, .R0 (B0 ), is to be
determined), .BZ > 0 and .0 < f < 1 modulation.
– an arbitrary field index .k—a good idea is to start building and testing the AVF in
the case .k = 0;
– a uniform map mesh in a polar coordinate system .(R, θ ) as sketched in Fig. 3.17,
covering R .= 1 to 100 cm; take a radial increment of the mesh .ΔR = 0.5 cm,
azimuthal increment.Δθ = 0.5 cm/RM , with.RM some reference radius, say.RM =
50 cm, half way between map boundaries;
– an appropriate 6-column formatting of the field map data for TOSCA to read, as
follows:
R cos θ, Z, R sin θ, BY , BZ, BX
.

with .θ varying first, .R varying second in that list. Z is the vertical direction (normal
to the map mesh), so .Z ≡ 0 in this 2D mesh.
Provide a graph of .BZ (R, θ ) over the extent of the field map.
(b) Raytrace a few concentric closed trajectories centered on the center of the dipole,
ranging in .10 ≤ R ≤ 80 cm. Provide a graph of these concentric trajectories in the
.(O; X , Y ) laboratory frame, and a graph of the field along trajectories. Initial coor-
dinates can be defined using OBJET, particle coordinates along trajectories during
the stepwise raytracing can be logged in zgoubi.plt by setting IL .= 2 under TOSCA.
(c) Check the effect of the integration step size on the accuracy of the trajectory
and time-of-flight computation, by considering some .Δs values in [0.1,10] cm, and
energies in a range from 200 keV to a few tens of MeV (considering protons).
(d) Produce a graph of the energy or radius dependence of wave numbers.
148 4 Relativistic Cyclotron

(e) Calculate the numerical value of the axial wave number, .νy , from the flutter
(Eqs. 4.5, 4.6). Comparing with the numerical values, discrepancy is found: repeat
(d) for f .= 0.1, 0.2, 0.3, 0.6, check the evolution of this discrepancy.

4.2 Designing an Isochronous AVF Cyclotron

Solution: 4.2.
(a) Introduce a radius dependent field index .k(R) in the AVF cyclotron designed
in Exercise 4.1, proper to ensure R-independent revolution period, in three different
cases of modulation: f .= 0 (no modulation), f .= 0.2 and f .= 0.9.
Check this property by computing the revolution period .Trev as a function of
kinetic energy .Ek , or radius .R. On a common graph, display both .Trev and .dTrev /Trev
as a function of radius, including for comparison a fourth case: B .= constant .= 5 kG.
(b) Provide a graph of the energy dependence of wave numbers.

4.3 Acceleration to 200 MeV in an AVF Cyclotron

Solution: 4.3.
In this exercise protons are accelerated to over 100 MeV in an AVF cyclotron:
well beyond the about 20 MeV energy reached in the classical cyclotron (see Exer-
cise 3.10).
(a) Produce an acceleration cycle of a proton, from 0.2 to 100 MeV, in the AVF
cyclotron designed in Exercise 4.2. Note that a dedicated field map has to be created
in order to allow for the higher maximum energy—a .3 m field map outer radius
works. Assume proper modulation coefficient .f for axial focusing all the way to
300 MeV. Assume a double-dee design, and 400 keV peak voltage in the gap, use
CAVITE[IOPT .= 7] for acceleration to account for RF phase.
(b) Give a graph of the energy dependence of wave numbers over the acceleration
range.

4.4 Thomas-BMT Spin Precession in Thomas Cyclotron

Solution: 4.4.
This exercise uses the field maps and input data file of Exercise 4.3. Dependence
of energy boost on RF phase is removed by using CAVITE[IOPT .= 3] [20]. Con-
sider helion ions: use PARTICUL[Name .= HELION] to define mass, charge and G
factor, all quantities needed for the integration of Thomas-BMT differential equation
(Eq. 3.30).
(a) By scanning the axial wave number, find the .Gγ value for which the spin
motion resonance condition (Eq. 4.18) is satisfied.
(b) Consider a particle with non-zero axial motion, so that it experiences horizontal
magnetic field components as it circles around. Track its spin through the resonance,
take initial spin vertical .S ≡ SZ . Provide a graph of .SZ as a function of .Gγ or energy.
(c) Simulate resonance crossing for a series of different vertical motion amplitudes
.Z0 ; produce a graph of these resonance crossings .SZ (turn).
Plot the ratio .Sy,f /Sy,i (Z0 ). From a match of this .Sy,f /Sy,i series with Eq. 4.19,
show that the resonance strength changes in proportion to the vertical excursion.
4.3 Exercises 149

(d) Repeat (c) for a series of different resonance crossing speeds instead (Eq. 4.20),
leaving .Z0 unchanged.
Show that this .Sy,f /Sy,i series can be matched with Eq. 4.19.

4.5 Isochronism and Edge Focusing in a Separated Sector Cyclotron

Solution: 4.5.
This exercise uses DIPOLE to simulate a 30 deg sector dipole of a 4-period
cyclotron, and allow playing with field fall-off extent at dipole EFBs. The configu-
ration of the cyclotron is typically that of PSI 72 MeV injector (Fig. 4.5). DIPOLE
allows radial field indices up to the third order (.∂ 3 BZ /∂R3 ) [20, Eq. 6.3.18]. In ques-
tion (b) however, higher order indices are needed to improve the isochronism, requir-
ing the use of DIPOLES [20, Eqs. 6.3.20, 21].
Take fringe fields into account (see Sect. 14.3.3), with

– .λ = 7 cm the fringe extent (changing .λ changes the flutter, Eq. 4.5),

– .C0 = 0.1455, .C1 = 2.2670, .C2 = −0.6395, .C3 = 1.1558 and .C4 = C5 = 0, for a
realistic field fall-off model.

(a) Assume .k = 0, here. Produce a model of a period using DIPOLE.

Produce a graph of closed orbits across a period for a few different rigidities (FIT
can be used to find them), and a graph of the field along these orbits.
(b) In this question, R-dependence of the mid-plane magnetic field proper to
ensuring energy independent revolution period is introduced. Use DIPOLES here,
as it allows .bi field indices to higher order, as necessary to reach tight isochronism
over the full energy range.
Assume a peak field value .B0 = 1.1 T at a radius of .3.5 m in the dipoles. Find the
average orbit radius R, and average field B (such that .BR = p/q), at an energy of
72 MeV.
Determine a series of index values, .bi=1,n , in the model [20, Eq. 6.3.19]
( ( )2 )
R − R0 R − R0
BZ (R, θ ) = B0 F (R, θ )
. 1 + b1 + b2 + ... (4.21)
R0 R0

proper to bring the revolution period closest to R-independent, in the energy range
0.9–72 MeV (hint: use a Taylor development of Eq. 4.15 and identify with the R-
dependent factors in Eq. 4.21).
(c) Play with the value of .λ, concurrently to maintaining isochronism with appro-
priate .bi values. Check the evolution of radial and axial focusing—OBJET[KOBJ
.= 5] and MATRIX[IORD .= 1, IFOC .= 11] or TWISS, or OBJET[KOBJ .= 6] and

MATRIX[IORD .= 2, IFOC .= 11], can be used to get the wave numbers.

150 4 Relativistic Cyclotron

From raytracing trials, observe that (i) the effect of .λ on radial focusing is weak (a
second order effect in the particle coordinates); (ii) greater (smaller) .λ value results
in smaller (greater) flutter and weaker (stronger) axial focusing (a first order effect).
Note: the integration step size in DIPOLE[S] has to be consistent with the field fall-off
extent (.λ value), in order to ensure that the numerical integration is converged.
(d) For some reasonable value of .λ (normally, about the height of a magnet gap,
⟨(B(θ)−<B>)2 ⟩
say, a few centimeters), compute .F 2 = . Check the validity of .νy =
√ <B>2
−β 2 γ 2 + F 2 (Eq. 4.6). OBJET[KOBJ .= 5] and MATRIX[IORD .= 1, IFOC .= 11]
can be used to compute .νy , or multiturn raytracing and a Fourier analysis.
hard edge
(e) Check the rule .F 2 −→ Rρ − 1 (Eq. 4.5), from the field .B(θ ) delivered by
DIPOLES. Give a theoretical demonstration of that rule.

4.6 A Model of PSI Ring Cyclotron Using CYCLOTRON

Solution: 4.6.
The simulation input data file in Table 4.2 is based on the use of CYCLOTRON, to
simulate a period of the eight-sector PSI ring cyclotron and work on the isochronism.
That file is the starting point of the present exercise.
(a) With zgoubi users’ guide at hand, explain the signification of the data in that
simulation input data file.
(b) Compute and plot a few trajectories and field along, across the sector. Provide
a graph of field density over the sector.
(c) Compute and plot the radius dependence of the revolution period.
(d) The field indices.b1 , b2 , ... are aimed at realizing the isochronism; four,.b1 − b4
are accounted for in (a) and (b), they were drawn from the PSI cyclotron spiral sector
magnet field map data. Question (c) proves this small set of indices to result in a poor
isochronism of the orbits.
Add higher order indices, until a sufficient number, with proper values, is found
that allows FIT to reach a final isochronism improved by an order of magnitude.
Provide a revised input data file with updated index series and their values.

4.4 Solutions of Exercises of This Chapter: Relativistic

Cyclotron

4.1 Modeling Thomas AVF Cyclotron

(a) A field map of a .360◦ AVF cyclotron dipole.

A Fortran program, geneAVFMap.f, given in Table 4.3, constructs the required map
of a field distribution .BZ (r, θ ), logged under the name geneAVFMap.out for use
by TOSCA keyword. A polar mesh is retained (Fig. 3.19), rather than Cartesian,
consistently with cyclotron magnet symmetry.
4.4 Solutions of Exercises of This Chapter: Relativistic Cyclotron 151

Table 4.2 Simulation input data file: a period of an eight-sector PSI-style cyclotron. The data file
is set up for a scan of the periodic orbits, from radius R.= 204.1171097 cm to R.= 383.7131468 cm,
in 15 steps

Note the following:

(i) The field map azimuthal extent (set to .360◦ in geneAVFMap, Table 4.3) can
be changed, for instance to simulate a 90 deg sector instead, with a sequence of four
simulating the complete ring.
(ii) Assuming mid-plane symmetry, the field in the (O;X,Y) plane is axial. The
field is taken radially constant in the first part of this exercise: .k = 0, thus .R(R) = 1
in Eq. 4.11.
(iii) The origin of the azimuthal angle in .BZ (R, θ ) = B0 [ 1 + f sin(4(θ − θi )) ] is
taken at .θi = π/2, leading to (Table 4.3)

.BZ (R, θ ) = B0 [ 1 + f cos(4θ ) ]

With this cosine dependence, and .θ covering .0 → 2π , the entrance and exit faces of
a .360◦ field map will be a location of maximum field (hill ridge), thus, owing to the
.2π/4 cylindrical symmetry of the field, the closed orbit is normal to these entrance
152 4 Relativistic Cyclotron

Table 4.3 A Fortran program, geneAVFMap.f, which generates a .360◦ mid-plane field map. This
angle as well as the field amplitude (.B0 = 5 kG at .R0 = 50 cm, here) and its modulation (.f = 0.2,
here) can be changed to any other values, a field index (.ak, set to zero here) can be accounted
for. The field map produced is logged in geneAVFMap.out, or under different names for the pur-
pose of the exercise, depending upon f, k, or AT values, e.g., geneAVFMap_90deg_f2_k0.out,
geneAVFMap_360deg_f9_k0.out...
4.4 Solutions of Exercises of This Chapter: Relativistic Cyclotron 153

and exit faces, so yielding TE .= TS .= 0 as KPOS arguments under TOSCA. The

same property holds in the case a 90 deg field map is used (.θ covering .0 → π/2):
take entrance and exit faces of the field map along hill ridges, i.e. normal to the closed
orbits.
As an indication of expected outcomes of this field map computation, the top
and bottom parts of the file generated by geneAVFMap, in the required format for
TOSCA[MOD.= 22.1] to swallow, are given in Table 4.3. Figure 4.4 displays the mid-
plane field so obtained (it uses the gnuplot script given at the bottom of Table 4.3)

(b) Concentric trajectories.

The input data file to raytrace trajectories with different rigidities (four, here) is given
in Table 4.4. The computation is performed in two steps
(1) in a first step, FIT finds the periodic coordinates for a given rigidity; note that
for this first step a 90 deg field map is INCLUDed (obtained with .AT = 90 deg in the
Fortran, Table 4.3; and used in subsequent exercises);
(2) upon completion of FIT, a second step computes the closed trajectory over
◦
.360 ; note: this double-step is one way to reduce the volume of zgoubi.plt file as it
is only written to at the second step, ounce the closed orbit has been found by FIT
(note: this can be accomplished in a single step, see Exercise 4.2). This process is
repeated for additional rigidities (i.e., additional orbits at different energies) using
REBELOTE[IOPT=1].
The following keywords are found in the input data file:
(i) OBJET: define a reference rigidity (arbitrary, taken to be 64.624444 kG cm,
here, corresponding to 200 keV protons), and define initial coordinates of a single
ion (initial radius .Y0 in the field map frame, other space coordinates zero; relative
rigidity D.= Bρ/BORO),
(ii) TOSCA: read the field map and raytrace; IL .= 2 flag causes log of particle
data in zgoubi.plt, after each integration step .Δs,
(iii) FAISCEAU: log local particle coordinates in zgoubi.res execution listing,
(iv) SYSTEM: run two gnuplot scripts once raytracing is completed, a first one
to plot the trajectories, a second one to plot the field along trajectories,
(v) MARKER: define two “LABEL_1” type labels, for use in INCLUDE state-
ments in subsequent exercises.
Four closed orbits resulting from the data file in Table 4.4 (for respective rela-
tive rigidities D .= 1, 1.5, 3.25, 5, spanning about 70 cm radially) are displayed in
Fig. 4.9. Inspecting zgoubi.res one finds the following particle coordinates as logged
by “’FAISCEAU’ CHECK” (Table 4.4), with initial values (left hand side) equal to
final values from the FIT procedure (right hand side), as expected:
154 4 Relativistic Cyclotron

Table 4.4 Simulation input data file FieldMapAVFMag.inc: raytrace a series of ions with different
rigidities, spanning 200 keV–5 MeV. This file also defines the optical segment #S_AVFMag_90d to
#E_AVFMag_90d, for use in subsequent exercises
4.4 Solutions of Exercises of This Chapter: Relativistic Cyclotron 155

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2
Ylab [m]

Ylab [m]
0 0
0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8

-0.2 -0.2

-0.4 -0.4

-0.6 -0.6

-0.8 -0.8
-0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8
Xlab [m] Xlab [m]

Fig. 4.9 Scalloping closed orbits in the 4-period AVF cyclotron with modulation factor .f = 0.2
(left) and .f = 0.9 (right)

The scalloping (orbit oscillation around the reference circle) is small, as can be
seen by comparison, below, with the closed orbit radius in the case of constant field.
The latter is obtained with a similar computation using a field map generated with
.f = 0; it can also be obtained from.R = p / qB with B.= 5 kG with.p = q × D × Bρref

and .Bρref = 64.6244440 kG cm the reference rigidity, under OBJET, yielding:

Figure 4.9 also displays an iteration of this closed orbits computation, yet for the
case of a modulation factor f .= 0.9 (thus using different field maps, named e.g.
geneAVFMap_90deg_f9_k0.out and geneAVFMap_360deg_f9_k0.out, for substi-
tution to the .f = 0.2 field map names in Table 4.4); the scalloping is increased due
to deeper modulation. Inspecting “’FAISCEAU’ CHECK” in zgoubi.res execution
listing one then finds the following particle coordinates for the 4 different rigidities:

The magnetic field along these orbits is displayed in Fig. 4.10, it is the same for
all four orbits as the field index is zero, here.
156 4 Relativistic Cyclotron

-4 -3 -2 -1 0 1 2 3 4
10

6
BZ [kG]

0
-4 -3 -2 -1 0 1 2 3 4
angle [rad]

Fig. 4.10 Four-periodic field .BZ (θ) along the closed orbits, case of a modulation factor .f = 0.9.
The field is the same ∫ for all four orbits as the field index is zero. The average value of the field along
a closed orbit is . π2 Δθ =π/2 BZ (R, θ) d θ = 5 kG

(c) Numerical convergence.

Numerical convergence of the stepwise integration is tested using the same input data
file as in (b) (Table 4.4, f .= 0.2 and k .= 0), the integration step size only, .Δs, needs be
changed, and the resulting change in accuracy translates in a change of closed orbit
coordinates as found by FIT. Two values of .Δs are tried (in addition to .Δs = 0.2 cm
in the previous computations, cf. Table 4.4). They yield the following outcomes of
FAISCEAU (at its occurrence at the bottom of Table 4.4, prior to REBELOTE), for
closed orbits at the four different relative rigidities D .= 1, 1.5, 3.25, 5:
. Case of .Δs = 1 cm
.
4.4 Solutions of Exercises of This Chapter: Relativistic Cyclotron 157

. Case of .Δs = 5 cm
.

The change of closed orbit coordinates is substantial for the lowest energy trajec-
tory, smaller circumference .C ≈ 2π × 13 ≈ 80 cm, covered in only 16 steps in the
case .Δs = 5 cm. Given the strong curvature, the high order derivatives of the field
vector take great values so jeopardizing the convergence of the position and velocity
vector Taylor series [20, Eq. 1.2.4]. The .Δs = 5 cm case features in addition poor
convergence of the FIT procedure, unable to zero the closed orbit angle in the small
radius cases, an effect of the field interpolation from a mesh.
(d) Dependence of wave numbers on energy and radius.
A scan of the wave numbers over a relative rigidity interval .D = BORO Bρ
: 1 → 5 is
performed using the input data file given in Table 4.5 (BORO is the reference rigidity,
under OBJET, D is the sixth coordinate of the reference particle as defined under
OBJET[KOBJ .= 5]). Wave numbers are computed using MATRIX.
OBJET[KOBJ .= 5] generates 13 particles with paraxial radial and axial coor-
dinates, and rigidity sampling, for the computation of transport matrix and wave
numbers by MATRIX. REBELOTE repeats this matrix computation sequence, for a
series of different rigidities. It is preceded by FIT which finds the closed orbit, this
is necessary as, (i) a different rigidity means different orbital radius, (ii) MATRIX
computes transport coefficients with respect to particle 1, which requires the latter
to be placed on the reference orbit, prior to MATRIX computation.
Inspection of the execution listing zgoubi.res shows the structure of a FIT at the
end of the FIT procedure, with the status of the variable (one variable only, here) in
a top block, followed by the status of the constraints in a bottom block. Here is an
excerpt of the FIT section in zgoubi.res, at the last iteration by REBELOTE (case of
relative rigidity D .= 5.00639):

Details regarding FIT[2] input, algorithms, and outcomes, are found in [20].
Further inspection of the execution listing shows the outcome of a MATRIX
command, under the form of two .6 × 6 blocks, a top one which is the transport
matrix .[Tij ] (see Sect. 14.5.2) from start to end of the optical sequence, and a bottom
one, “beam matrix” drawn from the periodicity hypothesis which allows to write (see
Sect. 14.5.2) .[Tij ] = I cos(μ) + J sin(μ). Here is an excerpt of the MATRIX section
158 4 Relativistic Cyclotron

Table 4.5 Simulation input data file: raytrace a set of 13 particles (defined by OBJET[KOBJ .=
5]) for a particular reference rigidity, to perform a MATRIX computation. FIT is used to find the
closed orbit, prior to MATRIX. Iteration for a series of 35 additional rigidities (relative rigidity D:
1.1.→5.00639, in 35 steps) is performed by REBELOTE. This input file INCLUDEs the segment
[#S_AVFMag_360d:#E_AVFMag_360d] of file FieldMapAVFMag.inc (Table 4.4)

in zgoubi.res execution listing, at the last iteration by REBELOTE (relative rigidity

D .= 5.00639):

The radial wave number .ν √ versus .1 − ν indetermination can be lifted by consid-

ering that .k = 0 so that .νR ≈ 1 + k ≈ 1, thus, actually, .νR = 1 − 0.054102699 =
0.945897301.
4.4 Solutions of Exercises of This Chapter: Relativistic Cyclotron 159

1.15 0.64345

1/2
1.1 0.6434

1.05 0.64335
(νR2+νy2)

2 2 1/2
(νR +νy )
νy

νy
νR

1 0.6433
νR,

0.95 0.64325

0.9 0.6432
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5
kin. E [MeV]

Fig. 4.11 A scan of the wave numbers as a function of proton energy in the cyclotron, with f .= 0.9
and k .= 0 here. Fluctuations stem from the use of a field map—performing the scan using DIPOLE
analytical field model instead, would yield smooth curves

MATRIX allows a PRINT command (Table 4.5), which causes the transport coeffi-
cients to be logged in zgoubi.MATRIX.out as REBELOTE iterates; reading from the
latter (gnuplot script given at the bottom of Table 4.5) √
yields Fig. 4.11. Results appear
reasonably close to theoretical approximations .νR ≈ 1 + k = 1, .νy ≈ F = 0.6364
and .(νR2 + νy2 )1/2 ≈ (1 + F 2 )1/2 = 1.185 (Eq. 4.6). The smaller the orbit scalopping
(modulation .f → 0), the better the agreement (see Table 4.6).

(e) Flutter.
√
The axial wave number writes (Eq. 4.8) .νy ≈ −k + F 2 = F. The flutter is given
( )1/2
by .F = <(F <F
−<F >)2 >
>2
(Eq. 4.5). The field modulation used here expresses as
.F = 1 + f cos N θ , and f .= 0.9. From this, one gets

∫ π/2
2
. < F >= (1 + f cos N θ ) d θ = 1
π 0

∫ π/2
2 f2
. < F >= (1 + f cos N θ )2 d θ = 1 +
2
π 0 2
( )1/2
< F 2 > − < F >2 f
F=
. = √ = 0.6364
< F >2 2

theoretical νy = F = 0.6364
.
160 4 Relativistic Cyclotron

Table 4.6 Wave number values in the case k .= 0, depending upon the field modulation f, from √
numerical raytracing (“ray-tr.” column) and from Eqs. 4.6, 4.7, namely,.νR = 1 and.νy = F = f / 2

Wave numbers
Radial, .νR axial, .νy .(ν
2+ νy2 )1/2
. R
√
f ray-tr. Eq. 4.6 ray-tr. .f / 2 ray-tr. .(1 +
f 2 /2)1/2
0.05 0.9999 1 0.0365 0.03535 1.0006 1.0006
0.1 0.9993 1 0.0730 0.0707 1.0020 1.0025
0.2 0.997 1 0.1459 0.1414 1.0076 1.0100
0.34 0.994 1 0.2185 0.2121 1.0177 1.0223
0.6 0.975 1 0.4338 0.4243 1.0671 1.0863
0.9 0.945 1 0.6433 0.6364 1.1432 1.1853

To assess wave numbers for different values of f, a series of field maps is to be

computed, as in (a), one for each f value. The outcomes of both numerical integra-
tion as in (d), and theoretical calculation as above, for different values of the field
modulation factor f, are summarized in Table 4.6. Discrepancy grows with greater
modulation, as Eq. 4.6 is a weak-modulation approximation [11, Sect. 3].

4.2 Designing an Isochronous AVF Cyclotron

(a) R-dependent field index.
A field index k(R) proper to ensure R-independent revolution period has to result in
(Eq. 4.14)

B0 M ωrev M ωrf
.B(R) = γ B0 = √ with B0 = = 2 (4.22)
1 − (R/R∞ )2 c 2 c h

For consistency with similar simulations in the Classical Cyclotron Chap. 3, the
following hypotheses are considered:
(i) injection energy .Einj = 200 keV,
(ii) average radius .Rinj = 0.129248888 m at that energy,
(iii) average field .Binj = B(R = Rinj ) = 0.5 T.
From this one gets .ωrev , the same at all R assuming isochronism, thus in particular

c2 Binj
.ωrev = = 2π × 7.62096882 × 106 rad/s wherein M γinj = M + 200 × 103
M γinj

with .M = 938.27208 × 106 eV /c2 , proton rest mass. In this exercise h .= 1 is

assumed, thus (Eq. 4.13)

c 2.99792458 × 108
R∞ =
. = = 6.2608118 m
ωrf 7.62096882 × 106
4.4 Solutions of Exercises of This Chapter: Relativistic Cyclotron 161
√
Using Eq. 4.22 the value .B0 ≡ B(R = 0) results, namely, .B0 = Binj 1 − (Rinj /R∞ )2
= 4.9989344, so, finally,

B0 4.9989344
B(R) = √
. =√
1 − (R/R∞ )2 1 − (R/6.2608118)2

The Fortran program geneAVFMapIsochro.f given in Table 4.7 constructs the map
for the.B(R, θ ) field distribution. It is derived from the Fortran program of Exercise 4.1
(Table 4.3) by accounting for the isochronism field dependence properties above. In
that file, the modulation factor .f can be changed, as well as the field index .k and the
angular extent of the field map, .AT . The resulting field distribution over 360 deg is
essentially as in Fig. 4.4 as the radial dependence of the field is weak: .B(R) = γ B0
whereas .γ ≈ 1, varying from .1.00000128 to .1.00745 over .R : 10 → 76 cm.
For the purpose of comparisons, four field maps are created and resorted to. Three
only differ by the value of the modulation coefficient (Table 4.7): .f = 0, 0.2, and
.0.9, an additional one is a “classical cyclotron” case (“Bcst” index, for constant

.B(R, θ )). In the latter case in addition

- BR = 1 is substituted to BR = 1/sqrt(1-(R/Rinfty)**2) and
- B0 = T2kG/2 is substituted to B0 = T2kG*Bp2k*sqrt(1-(Rp2k/Rinfty)**2).
In the following these field maps are handled under the following respective
names:
geneAVFMap_360deg_f0_isochro.out, geneAVFMap_360deg_f.2_isochro.out,
geneAVFMap_360deg_f.9_isochro.out and geneAVFMap_360deg_Bcst.out.
The input data file to raytrace ion orbits is given in Table 4.9. The FIT pro-
cedure finds the closed orbit for the particle defined by OBJET, REBELOTE
repeats for a series of additional rigidities in the range 1.1–5.×BORO (BORO .=
64.624444037 kG cm).
The exercise has been done for modulation factors f .= 0, 0.2, or 0.9, and as well
for a constant field .BZ (R, θ ) = 5 kG, as described above. The latter simulation shows
a great difference in the R dependence of the revolution time, compared to the two
isochronous cases. The sole cases .f = 0.2 and .f = 0.9 simulate an AVF cyclotron,
yielding stable axial motion; in the other two cases (f .= 0 and constant B) there is
no axial focusing, axial motion is unstable.
The resulting sets of closed trajectories are displayed in Fig. 4.12, the R depen-
dence of the revolution period in the four cases is given in Fig. 4.13. The revolution
period on the injection orbit for each of the four cases is given in Table 4.8.

(b) Wave numbers.

The energy dependence of wave numbers can be obtained by applying the procedure
of Exercise 4.1-d.

4.3 Acceleration to 200 MeV in an AVF Cyclotron

(a) Sufficient modulation has to be considered, for the axial focusing to be efficient
up to highest .γ (compensating the increase in .k(R)), namely (Eq. 4.8),
162 4 Relativistic Cyclotron

Table 4.7 A Fortran program, geneAVFMapIsochro.f, which generates an .AT = 360◦ mid-plane
field map of an isochronous cyclotron. .AT as well as the field amplitude (.B0 = 5 kG, here) and its
modulation (.f = 0.2, here) can be changed, a field index (.ak = 0, here) can be accounted for. The
field map produced is logged in geneAVFMapIsochro.out, it may be saved under a different name
for the purpose of the exercise, depending upon f, k, or AT values
4.4 Solutions of Exercises of This Chapter: Relativistic Cyclotron 163

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2
Ylab [m]

Ylab [m]
0 0
0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8

-0.2 -0.2

-0.4 -0.4

-0.6 -0.6

-0.8 -0.8
-0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8
Xlab [m] Xlab [m]

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2
Ylab [m]

Ylab [m]

0 0
0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8

-0.2 -0.2

-0.4 -0.4

-0.6 -0.6

-0.8 -0.8
-0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8
Xlab [m] Xlab [m]

Fig. 4.12 Twenty eight closed orbits in the field of a cyclotron. Top left: constant field .BZ = 5 kG;
top right: isochronous .B(R) field profile (Eq. 4.15) together with 4-periodic modulation (Eq. 4.3
with N .= 4) with f .= 0.9; bottom right: same, with f .= 0.2; bottom left: same, with f .= 0. The
f .= 0.9 and f .= 0.2 cases (right column) satisfy AVF focusing principles, the other two (case of
constant B and case f .= 0, left column) yield unstable optics due to the absence of axial focusing

√
.f > βγ 2

√ protons, up to over 100 MeV, i.e. .βγ > 0.474, axial focusing
Assume acceleration of
thus requires .f > βγ 2 = 0.67. A value of f .= 0.9 will be taken here.

This results in the 90 deg sector definition given in Table 4.10, which uses a field
map with a sufficiently large radial extent, geneAVFMap_90deg_f.9_isochro.out,
created using Table 4.7 program. Note that some cyclotron designs feature negative
valley field [21] to further increase the flutter (Eq. 4.5) and thus the axial focusing
(Eq. 4.6), so potentially allowing higher .k(R) and so higher energy (Eq. 4.1).
The voltage gap is simulated using CAVITE[IOPT=7]. Referring to Table 4.8 or
Fig. 4.14, the RF frequency has to be around 7.82 MHz, a little tweaking shows that
164 4 Relativistic Cyclotron

0.01

0.001 0.132

0.0001

0.131
1x10-5

1x10-6
ΔTrev/Trev

B=Cst

Trev [μs]
f=0 0.13
f=0.2
1x10
-7 f=0.9

1x10-8
0.129

-9
1x10

-10
1x10 0.128

-11
1x10
0.1 0.2 0.3 0.4 0.5 0.6 0.7
R [m]

Fig. 4.13 Left vertical scale, solid markers: departure from isochronism in the case of constant
field .BZ = 5 kG at all .(R, θ) (top curve; this is a “classical cyclotron” case) and (from bottom up)
of isochronous .B(R) (Eq. 4.14) with f .= 0, f .= 0.2 and f .= 0.9 (Eq. 4.3). Right vertical scale,
empty markers: revolution time; the “classical cyclotron” case (top curve) features steady increase
of revolution time due to mass increase

−1 ) at injec-
Table 4.8 Orbit length (.C), revolution period (.Trev ) and revolution frequency (.frev = Trev
tion, as a function of AVF modulation. Closed orbit length, and thus revolution period, tends to
decrease with increasing modulation
f .C (cm) .Trev (.µs) .frev (MHz)

Constant B 81.20948 0.13121691 7.6209688

0 81.20948 0.13121691 7.6209688
0.2 81.1014 0.13104242 7.6311163
0.9 79.12344 0.12784631 7.8218917

f = 7.7952 MHz yields efficient use of the RF. A 400 kV peak voltage is applied to
. rf

the electrode gap. This results in the input data file given in Table 4.11. Acceleration
cycles (and deceleration, beyond an RF phase of .π ) are shown in Figs. 4.15, 4.16.

(b) Energy dependence of wave numbers.

The energy dependence of wave numbers is displayed in Fig. 4.17 in the two modu-
lation cases f .= 0.1 and f .= 0.9. This simulation has been performed using the input
data file of Table 4.12. Two field maps have been generated for that purpose, using
the Fortran program in Table 4.7 with proper f values (the latter is the same as used
in Exercise 4.2). The argument PRINT under MATRIX causes logging of MATRIX
computation outcomes in zgoubi.MATRIX.out, including the wave numbers as plot-
ted in Fig. 4.17.
4.4 Solutions of Exercises of This Chapter: Relativistic Cyclotron 165

Table 4.9 Simulation input data file: raytrace a series of closed orbits with different rigidities,
spanning 200 keV to 5 MeV
166 4 Relativistic Cyclotron

Table 4.10 This file provides the simulation of a 90 degree AVF sector, with modulation f .= 0.9.
It defines an #S_AVFMag_90d_f9 to #E_AVFMag_90d_f9 segment subject to INCLUDE in the
input data file of Table 4.11. The END statement is mandatory at the end of an INCLUDE file

0.0001 7.8219

7.8218
1x10-5
7.8217

1x10-6 7.8216
ΔTrev/Trev

7.8215

frev [MHz]
ΔT/T
1x10-7 frev
7.8214

-8
1x10 7.8213

7.8212
-9
1x10
7.8211

-10 7.821
1x10
0.1 0.2 0.3 0.4 0.5 0.6 0.7
R [m]

Fig. 4.14 Left vertical scale, solid markers: departure from isochronism as a function of closed
orbit radius, case of f .= 0.9. Right vertical scale, empty markers: revolution frequency

The theoretical upper limit in energy, for axial stability, is determined by .βγ <
√
f / 2, i.e.,

– a theoretical 2.4 MeV for f .= 0.1, confirmed in this simulation, and

– a much higher 175 MeV for f .= 0.9, whereas this simulation yields 280 MeV (as
Eq. 4.6 is a weak modulation approximation).

4.4 Thomas-BMT Spin Precession in Thomas Cyclotron

Simulations use files developed in Exercise 4.3, with ad hoc modifications.

Helion is specified using PARTICUL. This determines the value of the gyromag-
netic anomaly, as well as mass and charge as they are needed to solve the differential
4.4 Solutions of Exercises of This Chapter: Relativistic Cyclotron 167

Table 4.11 Simulation input data file: acceleration gaps (two CAVITE) are added between two
180 deg sectors

Fig. 4.15 Acceleration Zgoubi|Zpop

24-07-2020 Kinetic energy (MeV) vs. pass #
followed by deceleration, 7.7952
case of f .= 0.9, for three 100 7.7962
different RF: 7.7942, 7.7952
and 7.7962 MHz
80

60

40 7.7942

20

50 100 150 200 250 300 350 400

168 4 Relativistic Cyclotron

Y (m) vs. pass # Zgoubi|Zpop Z (m) vs. pass #

24-07-2020 24-07-2020
0.02
7.7952
7.7962 0.015
2.5
0.01
2.
0.005

1.5 0.0

7.7942 -.005
1.
-.01

0.5 -.015

50 100 150 200 250 300 350 400 50 100 150 200 250 300 350 400

Fig. 4.16 Left: radial excursion during acceleration and deceleration, case of f .= 0.9, for the three
different RF 7.7942, 7.7952 and 7.7962 MHz. Right: axial excursion, case of .frf = 7.7952 MHz

Tune dependence on energy, from zgoubi.MATRIX.out Tune dependence on energy, from zgoubi.MATRIX.out
1.0035 1.1 1.3 1.4
1
1.2
1.003 0.9
1.25
0.8 1
νy, νR2+νy2 - f2/2

νy, νR2+νy2 - f2/2

0.7
1.0025 νR
1.2 νR 0.8
0.6
νR

νy νy
νR

2 2 2
νR +νy -f /2
0.5 2 2 2
νR +νy -f /2 0.6
1.002 1.15
0.4
0.3 0.4
1.0015 0.2 1.1
0.2
0.1
1.001 0 1.05 0
0 0.5 1 1.5 2 2.5 3 0 50 100 150 200 250 300
kinetic E [MeV] kinetic E [MeV]

Fig. 4.17 The left and right graphs are for respectively f .= 0.1 and f .= 0.9 modulation factor. Left
vertical scale: radial wave number. Right vertical scale: axial wave number and .νR2 + νy2 − f 2 /2,
the latter expected constant and close to 1 in the small scalloping/weak modulation approximation
(Eq. 4.8). The upper limit in energy is determined by.νy decreasing to zero, namely, around 2.4 MeV
for f .= 0.1, around 280 MeV for f .= 0.9

equation of spin motion (Eq. 3.30). PARTICUL results in the following print out in
zgoubi.res execution listing:
4.4 Solutions of Exercises of This Chapter: Relativistic Cyclotron 169

Table 4.12 Simulation input data file: energy dependence of wave numbers. The INCLUDE uses
the TOSCA segment defined in Table 4.10

(a) Resonant .Gγ .

A preliminary scan of motion wave numbers is performed, using the input file of
Table 4.13. This scan shows that, over a kinetic energy range .Ek : 50 → 300 MeV,
the axial wave number .νZ decreases from 0.6 to 0.25 about while .Gγ : −4.25 →
−4.65 (Fig. 4.18). It results that at a particular location over that energy range, the
relationship

. Gγ + νZ = integer = −4

is satisfied, namely here: .Gγ = −4.4375.

(b) Helion spin precession.

A spin tracking is launched with the helion ion injected near the.Bρ = 64.624444 T m
closed orbit, namely, .Rinj ≈ 13 cm and angle .Tinj = 0, with non-zero axial
170 4 Relativistic Cyclotron

Table 4.13 Simulation input data file: a scan of wave numbers, computed using OBJET[KOBJ
.= 5] and MATRIX, in 74 steps over a relative rigidity range .D : 1 → 36, i.e., helion rigidity
.Bρ : 64.624444 → 36 × 64.624444 T m, energy.E : 0.267292 → 2.326479 MeV. The INCLUDE
uses the TOSCA segment defined in Table 4.10. FIT finds particle closed orbit and spin .n0 vector,
prior to MATRIX computation

amplitude in order to excite the spin resonance, namely, .Zinj = 2 cm (vertical take-
off angle .Pinj = 0). The simulation file is given in Table 4.14. Acceleration is over
. Gγ : −4.18 → −4.75, the axial wave number decreases from 0.64312 to 0.23011.
Figure 4.18 displays the vertical spin component, flipping from .+1 to .−1; a close
inspection of raytracing outcomes confirms the location of the resonance at .GγR =
−4 − 0.4375.

(c) Spin resonance crossings. Resonance strength.

This exercise is performed by repeating the simulation of Table 4.14 for a series of
different .Z0 values; outcomes are displayed in Fig. 4.19.
As expected (Eq. 4.19) .SZ,f /SZ,i tends toward 1 (respectively, toward .−1), as the
strength of the resonance tends toward zero (respectively, goes .≫ 0), Fig. 4.20. The
4.4 Solutions of Exercises of This Chapter: Relativistic Cyclotron 171

Table 4.14 Simulation input data file: spin tracking through the .Gγ + νZ = −4 resonance

kinetic E [MeV]
50 100 150 200 250 300
0.65 1

0.8
0.6

0.6
0.55
0.4

0.5
0.2
νZ, |Gγ|-4

SZ
νZ
SZ

0.45 0
|Gγ|-4
-0.2
0.4

-0.4
0.35
-0.6

0.3
-0.8

0.25 -1
4.25 4.3 4.35 4.4 4.45 4.5 4.55 4.6 4.65
|Gγ|

Fig. 4.18 Spin resonance crossing. The graph shows the evolution of the axial wave number .νZ
and of the quantity .|Gγ | − 4 (left vertical axis), and of the helion ion spin, initially vertical, .SZ = 1
(right vertical axis), as a function of .Gγ (lower horizontal axis) and of energy (upper horizontal
axis). .νZ and .|Gγ | − 4 curves cross at .Gγ = −4.4375
172 4 Relativistic Cyclotron

0.5
SZ

-0.5

-1
0 500 1000 1500 2000 2500 3000 3500 4000
Turn

Fig. 4.19 Evolution of .SZ during resonance crossing, for a series of values of the initial axial
particle coordinate .Z0 . Spin flip occurs at larger .Z0 values

1
raytracing
theory

0.5
SZ,f/SZ,i

-0.5

-1
0 0.5 1 1.5 2
Z0

Fig. 4.20 Evolution of .SZ,f /SZ,i toward spin flip as the axial motion excursion increases, from
raytracing (markers) and from theory (Eq. 4.19 with .|ϵR | ∝ Z0 )
4.4 Solutions of Exercises of This Chapter: Relativistic Cyclotron 173

former case corresponds to absence of resonance, i.e., .BZ axial always, as .Z0 ≡ 0:
the ion motion is in the median plane of the cyclotron dipole. Increasing .Z0 increases
the strength of the non-vertical field experienced by the ion as it cycles around the
accelerator, and causes spins to undergo greater tilt at traversal of the resonance,
toward spin flip with sufficient vertical excursion.
A match of .SZ,f /SZ,i (Z0 ) to Eq. 4.19 shows that these raytracing outcomes satisfy
.|ϵR | ∝ Z0 .

(d) Changing the crossing speed.

The method to answer this question is the same as in (c), repeating the simulation of
Table 4.14 for a series of different acceleration rates (i.e. gap voltage, V) in CAVITE,
to get .Sy,f /Sy,i (V ).
Finally, the relationship to the crossing speed (Eq. 4.20 ) can be established using
the .d νZ /dt data produced in (b) (Fig. 4.18).

4.5 Isochronism and Edge Focusing in a Separated Sector Cyclotron

A separated sector isochoronous cyclotron modeled using DIPOLE.

(a) DIPOLE allows to account for the field fall-off extent at dipole EFBs, which
determines the flutter. The input data file for this simulation is given in Table 4.15.
Across the 30 deg sector dipole, a 4-periodic closed orbit undergoes a 90 deg bend,
whatever the rigidity. Due to the periodicity and to the field symmetry (the dipole is
symmetric with respect to a vertical plane at 15 deg to its EFBs), the closed orbits
enter and exit the magnetic sector with angles .TE = TS = 0.

FIT is used to find the closed orbit at a particular rigidity, the process is repeated
(using REBELOTE[IOPT .= 1]) for a series of different rigidities, in the following
way:
– the first constraint under FIT imposes that particle 1 be on a periodic orbit. That
constraint is enforced with a weight of 0.1, i.e. greater compared to 1 for the second
constraint;
– for that, FIT allows varying .B0 , and ends up with the same .B0 always, as expected
given .k = 0. This first constraint is maintained unchanged during the REBELOTE
process (which repeats with a different rigidity, yet first changing the relative
rigidity D of the second particle - D datum at position 45 in OBJET);
– the second constraint concerns the radial coordinate of closed orbits: it requires
that the initial Y coordinate (Y coordinate at OBJET) of particle 1, be equal to its
final coordinate (after DIPOLE), a closed orbit condition (Figs. 4.21 and 4.22).

(b) Isochronous B(R).

A similar problem is treated in Exercise 4.6, thus just indications are given here, as
to determining a proper radial field law for isochronism.
174 4 Relativistic Cyclotron

Table 4.15 Simulation input data file 90degEdgeFocusSector.inc: analytical modeling of a

30 degree magnetic sector of a 4-period separated sector cyclotron. This simulation file includes
a search of cyclotron orbits for eight different energies in .[0.87, 72] MeV. The LABEL_1s
#S_90degCycloSector and #E_90degCycloSector define the dipole segment, for further use in
subsequent exercises
4.4 Solutions of Exercises of This Chapter: Relativistic Cyclotron 175

2.5

1.5
Ylab [m]

0.5

0
0 0.5 1 1.5 2 2.5 3
Xlab [m]

Fig. 4.21 Closed obits across a quadrant, at a few different rigidities, from raytracing

14

12

10

8
Ylab [m]

-2
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
Xlab [m]

Fig. 4.22 Field along closed obits at different rigidities, over a quadrant, from raytracing
176 4 Relativistic Cyclotron

R0 ∂B
The indices in Eq. 4.21 can be expressed under the form .b1 = B0 ∂R
, b2 =
R20∂2B R30∂3B
2B0 ∂R2
,b3 = etc. Expand the .(R − R0 ) terms in Eq. 4.21 and re-organize in
6B0 ∂R3
, i

increasing powers of R, so writing the radial dependence of the field under the form

. R(R) = (1 − b1 + b2 − b3 + b4 + ...) + RR0 (b1 − 2b2 + 3b3 − 4b4 + ...)

( )2 ( )3
+ RR0 (b2 − 3b3 + 6b4 + ...) + RR0 (b3 − 4b4 + ...) + ... (4.23)

On the other hand, the Taylor series development of the R-dependent factor of the
magnetic field for isochronism, Eq. 4.14, writes

1 (R/R∞ )2 3(R/R∞ )4 5(R/R∞ )6

R(R) ≈ /
. =1+ + + + ... (4.24)
R 2 2 8 16
1−( )
R∞

Identify term by term with Eq. 4.23, this yields the indices .bi in terms of powers
of .1/R0 (.R0 is a known quantity), the very values to be used in defining the field
and indices in DIPOLES. Accuracy on isochronism can be improved using FIT[2]:
require isochronism (the constraint in FIT[2]) and allow varying the .bi indices in
DIPOLES (the variables in FIT[2]) starting from initial values obtained as described
above.

(c) Changing field fall-off extent.

Indications:
Changing the fringe field extent .λ impacts both the closed orbit landscape and
the isochronism. The latter can then be re-optimized by means of FIT, varying the
.bi coefficients and constraining, concurrently and over the energy extent of concern,
both the orbit periodicity and the isochronism of these orbits. Such a FIT is performed
in Exercise 4.6, the same method can be applied here.

(d, e) Flutter, axial wave number.

Indications:
Graphs of .R− or .βγ -dependence of wave numbers, and relationship to the flutter,
are produced in Exercise 4.1, the same techniques can be applied here.

4.6 A Model of PSI Ring Cyclotron Using CYCLOTRON

CYCLOTRON provides a realistic analytical modeling of the field in a radial or spiral
sector magnet of a separated sector cyclotron [20, Sect. 6.3 & Part B]. CYCLOTRON
keyword belongs in the DIPOLE[S] and FFAG[-SPI] families, with some specifici-
ties. The large number of field indices available is one, it simulates pole shaping and
allows fine tuning of the isochronism.
4.4 Solutions of Exercises of This Chapter: Relativistic Cyclotron 177

300

200

100
dexit
. P (R,θ)
Y (cm)

dentr
0 AT

Spiral boundaries
-100

-200

-300
0 100 200 300 400 500 600
X (cm)

Fig. 4.23 A representation of the EFBs in CYCLOTRON, encompassing a sector of AT angle.

The value of the flutter .F (R, θ) at ion location .P(R, θ) is determined from the distance .d to the
EFBs. If there are several dipoles within AT, all EFBs are accounted for, in computing the field at
.P(R, θ) [23]

(a) CYCLOTRON data list.

A sketch of a PSI cyclotron spiral sector as simulated here, and corresponding to
CYCLOTRON input data list of Table 4.2, is given in Fig. 4.23. A commented version,
in answer to question (a), is given in Table 4.16. A note on the origin of the data used
to simulate a cyclotron sector, in Table 4.2:
(i) The parameters needed in the equation of the spiral effective field bound-
aries [20, Eq. 6.3.15], and to determine the effective magnetic field length, have been
obtained using the magnetic field map of PSI cyclotron [22]. In particular

• at entrance EFB: .ω = 11 deg and .ξ [deg] = 3.5 + 35.10−3 r + 3.10−8 r 3 ;

• at exit EFB: .ω = −8.5 deg and .ξ [deg] = 2 + 12.10−3 r + 75.10−6 r 2 .

(ii) The radial field law .R(R) has been obtained by fitting the field fall-off along a
series of closed orbits at different radii in the magnetic field map, which yielded the
polynomial coefficients .b0 to .b4 . A fitting of the 137 MeV closed orbit in particular
provided the fringe field coefficients .C0 to .C5 .
178 4 Relativistic Cyclotron

Table 4.16 Simulation input data file: a period of PSI eight-sector CYCLOTRON model. The data
file is set up for a scan of the closed orbits, from radius R.= 204.1171097 cm to R.= 383.7131468 cm,
in 15 steps. Comments have been added, line by line, as a guidance
4.4 Solutions of Exercises of This Chapter: Relativistic Cyclotron 179

0.2 16
Orbit used to fit the Fringe field coeff Fieldmap
Other orbits CYCLOTRON
0.15 14

0.1 12
Ranal-Rmap/Rmap (%)

0.05 10

Bz (kG)
0 8
240 MeV
-0.05 6

-0.1 4

-0.15 2
137 MeV
-0.2 0
0 50 100 150 200 250 300 0 50 100 150 200 250 300
Path length (cm) Path length (cm)

Fig. 4.24 Left: relative difference in radial excursion across a cell, for a few closed orbits (at 137,
156, 176, 196, 218 and 240 MeV), between raytracing in the analytical model CYCLOTRON, and
in PSI field map [22]. Right: field profiles along these orbits, using indifferently the analytical model
CYCLOTRON, or PSI field map: differences are not noticeable at this scale

(b) Pole and field profiles, closed orbits.

The CYCLOTRON input data file, Table 4.16, can be run for various particle rigidities
(change D in the particle coordinates under OBJET[KOBJ=2]), possibly with sev-
eral particles (OBJET[KOBJ=2,IMAX>1]). IL=2 under CYCLOTRON takes care
of storing stepwise particle data in zgoubi.plt, to produce graphs.
Outcomes are illustrated in Fig. 4.24 which compares closed orbit excursions,
whether using the analytical CYCLOTRON model, or PSI field map from which it
is drawn [22].
Closed orbits at a large number of different rigidities can be raytraced (use
OBJET[KOBJ=2,IMAX.≫1]), from which a graph of isomagnetic field lines can
be produced, by reading from zgoubi.plt. This allows producing Fig. 4.25 in which
the EFBs and a series of closed orbits are superposed on a field scale background.
Note: the closed orbit for a particular rigidity can be found using FIT, the process
can be repeated using REBELOTE[IOPT=1] (as in Table 4.16).

(c) Revolution period.

The Simulation input data file of Table 4.16 performs the scan needed here, comments
therein explain the method, which is based on FIT to find the proper rigidity and
periodic orbit angle for a particle with periodic radius defined by OBJET, and on
REBELOTE[IOPT=1] to repeat the FIT procedure for a new value of the orbit radius,
NPASS times.
Figure 4.26 checks the proper completion of the FIT procedure, showing that
final orbit radius .R = Y (down the magnetic sector) is identical to initial .R = Y0 (at
OBJET) and final orbit angle T is identical to initial .T0 . Note that a global check is
provided by the penalty value, an outcome also of the FIT procedure.
Figure 4.27 displays the relative time difference to that of a reference orbit, taken
to be orbit number 7 at R .= 314 cm in the middle region of the range, bottom of the
time of flight parabola.
180 4 Relativistic Cyclotron

Bz (kG)
200 25
Closed Orbits obtained from the fieldmap
150 EFB at the entrance of the magnet
EFB at the exit of the magnet 20
C
100
15

50
Y (cm)

10
B
0

5
-50

0
-100

A
-150 -5
150 200 250 300 350 400 450 500
X (cm)

Fig. 4.25 EFBs and field scale, obtained from raytracing using CYCLOTRON or, indifferently,
PSI sector field map [22]. B is the location of the maximum field value along the 137 MeV orbit, C
is the location of the minimum value in the field valley

T0 [mrad]
-100 -90 -80 -70 -60 -50 -40 -30 -20 -10
360 -10
’orbits.fai’ u colY0:colY
350 ’orbits.fai’ u colT0:colT -20
340 -30
330
-40
320
T [mrad]

-50
Y [cm]

310
-60
300
-70
290

280 -80

270 -90

260 -100
260 270 280 290 300 310 320 330 340 350 360
Y0 [cm]

Fig. 4.26 Checking the proper completion of the FIT procedure: final orbit radius .R = Y (down
the magnetic sector) is identical to initial .R = Y0 (at OBJET). The constraint is similar for orbit
angles: final orbit angle .T identical to initial .T0
4.4 Solutions of Exercises of This Chapter: Relativistic Cyclotron 181

0.006
’orbits.fai’ u colY0:(($15-T314)/T314)

0.005

0.004
(Trev-TR=314)/TR=314

0.003

0.002

0.001

-0.001
260 270 280 290 300 310 320 330 340 350 360
R [cm]

Fig. 4.27 Time of flight difference as a function of closed orbit radius, relative to time of flight
.T (R= 314.46264 cm) = 0.01473792 µs

(d) Improved isochronism.

The number of field indices accounted for in .R(R) (Eq. 4.21) in setting up
CYCLOTRON modeling (four only in the case of questions (a) and (b), .b1 to .b4 ) is
increased iteratively, one additional index at a time, and the FIT procedure is re-run
each time, until it shows convergence to a series of index values which result in the
required degree of isochronism. An additional eight indices allow an improvement
of the isochronism by a factor 50. The main field .B0 is also part of the variables,
as it allows the orbits to adjust to periodic condition (closed orbits). The simulation
data file is given in Table 4.17, including the FIT procedure at the last stage of the
iteration on the number of field indices.
Note in the FIT procedure constraints: the time of flight on orbit 4, middle
region of the range, bottom of the time of flight parabola, is taken as a reference.
It does not act as a constraint in the FIT as its weight is .109 , compared to .10−4
and less for the other 6 orbits. That orbit 4 ends up reaching the revolution time
value .Tref = 0.0147442157 μs.
The resulting relative time of flight difference .dTrev /Tref = (Trev − Tref )/Tref as a
function of closed orbit radius is displayed in Fig. 4.29, much improved by compar-
ison to the 4 index case (Figs. 4.27 and 4.28).
182 4 Relativistic Cyclotron

Table 4.17 Simulation input data file: a period of PSI separated sector cyclotron, using
CYCLOTRON for an analytical modeling of the field. The file is setup to FIT 12 field indices,
.b1 to .b12 for improved isochronism. The constant field .B0 is part of FIT variables, to allow for the
constraint of orbit periodicity. OBJET[KOBJ .= 2, IMAX .= 7] creates 7 particles which span the
momentum range of interest, via .D : 1.4 → 2.1
4.4 Solutions of Exercises of This Chapter: Relativistic Cyclotron 183

T0 [cm]
-100 -90 -80 -70 -60 -50 -40 -30 -20 -10
360 -10
Y(Y0)
350 T(T0) -20
340 -30
330
-40
320
-50
310
Y

T
-60
300
-70
290

280 -80

270 -90

260 -100
260 270 280 290 300 310 320 330 340 350 360
Y0 [cm]

Fig. 4.28 Checking the proper completion of the 12-index FIT procedure: the final orbit radius
.R= Y (down the magnetic sector) is identical to the initial .R = Y0 (at OBJET). Same constraint
for orbit angles: the final orbit angle T comes out identical to the initial .T0

-5
0.006 4x10
4-index
12-index
’’ zoom
0.005 2x10-5

0.004 0
dTrev/Tref (zoomed)

0.003 -2x10-5
dTrev/Tref

0.002 -4x10-5

-5
0.001 -6x10

0 -5
-8x10

-0.001 -0.0001
260 270 280 290 300 310 320 330 340 350 360
perioric orbit radius, R [cm]

Fig. 4.29 A graph of the improved isochronism with 12 field indices (circles and left vertical axis,
and a zoom-in: triangles and right axis; data are read from the file FITted.fai), compared to results
obtained in question (b) (squares) where 4 indices were used. The isochronism is improved by a
factor of ∼50
184 4 Relativistic Cyclotron

References

1. A link to accelerator conference proceedings and its search tool: https://www.jacow.org/Main/

Proceedings
2. J.M. Schippers, The superconducting cyclotron and beam lines of PSI’s new proton therapy
facility “PROSCAN”, in 17th International Conference on Cyclotrons and their Applications,
Tokyo, Japan (2004). Accessed from 18–22 Oct 2004. http://accelconf.web.cern.ch/c04/data/
CYC2004_papers/20B2.pdf. Figure 4.1: https://accelconf.web.cern.ch/cyclotrons2019/talks/
tub04_talk.pdf Copyrights under license CC-BY-3.0, https://creativecommons.org/licenses/
by/3.0; no change to the material
3. M. Seidel, Production of a 1.3 megawatt proton beam at PSI, in Proceedings of
IPAC’10, Kyoto, Japan, TUYRA03. http://accelconf.web.cern.ch/IPAC10/talks/tuyra03_talk.
pdf. Figure 4.2: https://indico.psi.ch/event/3484/attachments/5948/7494/Cyclotrons_PSI.pdf
Copyrights under license CC-BY-3.0, https://creativecommons.org/licenses/by/3.0; no change
to the material
4. T. Kawaguchi et al., Design of the sector magnets for the RIKEN superconducting ring
cyclotron, in Proceedings of the 15th International Conference on Cyclotrons and their Appli-
cations, Caen, France (2003). https://accelconf.web.cern.ch/c98/papers/b-14.pdf. Figure 4.3:
the picture, and the permission to use it in this publication, have been granted by RIKEN
5. L.H. Thomas, The paths of ions in the cyclotron. Phys. Rev. 54, 580 (1938)
6. W. Joho, Potential of cyclotrons. https://indico.in2p3.fr/event/115/timetable/?print=1&
view=standard
7. L.H. Thomas, Motion of the spinning electron. Nature 117(2945), 514 (1926)
8. K.R. Symon et al., Fixed-field alternating-gradient particle accelerators. Phys. Rev. 103, 1837
(1956)
9. A.I. Yavin, The AVF cyclotron. Phys. Today 15(5), 19–25 (1962). https://doi.org/10.1063/1.
3058175
10. L.B. Cohen, Cyclotrons and synchrocyclotrons, in Encyclopedia of Physics, ed. by S. Flūgge,
Vol. XLIV, Nuclear Instrumentation I. (Springer, 1959)
11. T. Stammbach, Introduction to cyclotrons. CERN accelerator school, cyclotrons, linacs and
their applications, IBM International Education Centre, La Hulpe, Belgium (1994). Accessed
from 28 April–5 May 1994. Copyright/License CERN CC-BY-3.0. https://creativecommons.
org/licenses/by/3.0
12. P.S. Miller et al., The magnetic field of the K500 cyclotron at MSU including trim coils and
extraction channels, in Proceedings of 9th International Conference on Cyclotrons and their
Applications, September 1981, Caen, France. http://accelconf.web.cern.ch/c81/papers/ep-05.
pdf
13. H.A. Bethe, M.E. Rose, Phys. Rev. 54, 588 (1938)
14. F.T. Cole, A memoir of the MURA years (1994). Accessed from April 11 1994. https://epaper.
kek.jp/c01/cyc2001/extra/Cole.pdf
15. © The Regents of the University of California, Lawrence Berkeley National Laboratory
16. M.K. Craddock, AG focusing in the Thomas cyclotron of 1938, in Proceedings of PAC09,
Vancouver, BC, Canada, FR5REP1. http://accelconf.web.cern.ch/PAC2009/papers/fr5rep113.
pdf
17. F. Méot, Spin dynamics, in Polarized Beam Dynamics and Instrumentation in Particle Acceler-
ators, USPAS Summer 2021 Spin Class Lectures. Springer Nature, Open Access (2023). https://
link.springer.com/book/10.1007/978-3-031-16715-7
18. F. Méot, Spinor methods, in Polarized Beam Dynamics and Instrumentation in Particle Accel-
erators, USPAS Summer 2021 Spin Class Lectures. Springer Nature, Open Access (2023).
https://link.springer.com/book/10.1007/978-3-031-16715-7
19. M. Froissart, R. Stora, Depolarisation d’un faisceau de protons polarises dans un synchrotron.
Nucl. Instrum. Methods. 7(3), 297–305 (1960). (June)
References 185

20. F. Méot, Zgoubi Users’ Guide. https://www.osti.gov/biblio/1062013-zgoubi-users-guide.

Sourceforge latest version: https://sourceforge.net/p/zgoubi/code/HEAD/tree/trunk/guide/
Zgoubi.pdf
21. P. Mandrillon, Single Stage Cyclotron for an ADS Proceedings of Cyclotrons 2016, Zurich,
Switzerland. http://accelconf.web.cern.ch/cyclotrons2016/talks/fra01_talk.pdf
22. M. Haj Tahar, Tutorial - Case Study: PSI Cyclotron, using CYCLOTRON and TOSCA com-
mands in Zgoubi. Zgoubi and OPAL Users Mini-workshop, in FFAG’14 Workshop, BNL
(2014). https://www.bnl.gov/ffaworkshop/events/index.php#2014
23. F. Lemuet, F. Méot, Developments in the raytracing code Zgoubi for 6-D multiturn tracking in
FFAG rings. NIM A 547, 638–651 (2005)

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0
International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing,
adaptation, distribution and reproduction in any medium or format, as long as you give appropriate
credit to the original author(s) and the source, provide a link to the Creative Commons license and
indicate if changes were made.
The images or other third party material in this chapter are included in the chapter’s Creative
Commons license, unless indicated otherwise in a credit line to the material. If material is not
included in the chapter’s Creative Commons license and your intended use is not permitted by
statutory regulation or exceeds the permitted use, you will need to obtain permission directly from
the copyright holder.
Chapter 5
Betatron

Abstract This chapter introduces the betatron fixed orbit cyclic accelerator. It begins
with a brief reminder of the historical context, and continues with the Widerøe con-
dition and the principles of fixed orbit acceleration in a betatron. The latter is at the
origin of the theory of the “betatron oscillations”—treated in Chaps. 3 and 8. A realis-
tic simulation of a betatron in zgoubi would require the simulation of an induction
electric field: this is doable from existing dipole models such as DIPOLE[S], and can
be seen as an interesting code development exercise. A simpler approach on the other
hand only requires two optical elements: DIPOLE and CAVITE. Accounting for syn-
chrotron radiation (SR) energy loss requires SRLOSS. Monte Carlo SR monitoring
can use SRPRNT, which logs data in zgoubi.res. SRPRNT[PRINT] in addition logs
data in zgoubi.SRPRNT.Out. Electron beam monitoring requires keywords intro-
duced in the previous chapters, such FAISCEAU, FAISTORE. SR monitoring uses
SRPRNT. INCLUDE allows simplifying the input data files. Graphs are part of data
treatment and simulation outcomes, they are produced using zpop or gnuplot.

Notations Used in the Text

.A; As,r,y vector potential, .curlA = B; its components

.B; Bs,r,y ; B̂ magnetic field vector; its components; peak value
. Ba average value of magnetic field circumscribed by the orbit
.D dispersion, . D = δx / δp/ p
. E; Ê electron energy, . E = γ m 0 c2 ; peak value
.e elementary charge
. Ea induction accelerating field
.E; E s,r,y induction electric field vector; its components
.m; .m 0 ; . M electron mass; rest mass; in units of MeV/c.2
ρ ∂B
.n = − focusing index
B ∂x
.p electron momentum
.R radius of equilibrium orbit

© The Author(s) 2024 187

F. Méot, Understanding the Physics of Particle Accelerators, Particle Acceleration
and Detection, https://doi.org/10.1007/978-3-031-59979-8_5
188 5 Betatron

.s path variable
.Us synchrotron radiation energy loss [ ]
, ,
.x , y radial and axial coordinates in the moving frame. (∗), = d(∗)
ds
.β = v/c normalized velocity
.βu betatron functions (.u : x, y)
.γ = E/m 0 c
2
Lorentz relativistic factor
.δp, ./\p momentum offset
.Ec critical energy of SR, .Ec = hωc = hc/λc
.εu Courant–Snyder invariant (.u : x, y)
.νu wave numbers, radial, vertical (.u : x, y)
.ω angular frequency of the cycling field
.ωc critical angular frequency of SR, .ωc = 3γ 3 c/2ρ
.ωrev angular revolution frequency
.ωR , ωy betatron frequency, radial, axial
.φa magnetic flux

5.1 Introduction

The concept of an inductive electric field accelerating electrons, maintained on a

constant orbit in a magnetic field, goes back to the early 1920s [1, 2]. That principle
of inductive acceleration has been in use over the years in a variety of systems, for
instance in synchrotrons for slow extraction [3]; in ion synchrotrons as the acceler-
ation system proper [4]; in induction linacs [5, 6]; in fixed field alternating gradient
accelerators (FFAG) [7–9] (Fig. 5.1). The betatron ring is based on this acceleration
technique, it is an induction accelerator. The development of the early 1920s con-
cept, toward success in 1940, was fostered by the need for high energy electrons and
for X-rays. The earlier classical cyclotron was of no help, it only works for under-
relativistic particles, which for electrons means keVs. And five more years would be
needed for phase focusing to revolutionize cyclic acceleration.
Betatron acceleration on a constant orbit requires Widerøe’s condition, estab-
lished by the latter in the late 1920s [1]. Transverse confinement in addition requires
focusing. Operation of a betatron ring, bringing these two prerequisite together, was
first achieved in 1940, with the acceleration of electrons to 2.3 MeV, on the rising
slope of a several 100 Hz magnetic field impulse [10, 11] (Fig. 5.2). This was the
first cyclic accelerator, and the first electron ring. The energy reach of the technology
culminated with a 315 MeV betatron at the University of Chicago, in the late 1940s,
with an orbital radius of 1.2 m for a guiding field strength close to 1 T, requiring
a 350 ton magnet. At that energy, synchrotron radiation starts causing significant
energy loss, which can be somewhat mitigated with additional RF power, yet will
eventually jeopardize Widerøe’s condition for a fixed orbit. This was a limitation for
higher energy. The betatron on the other hand is inefficient to accelerate ions from
5.1 Introduction 189

Fig. 5.1 Induction acceleration in (left) the MURA 180 keV spiral sector FFAG electron model,
circa 1957 [7], and in (right) the Ion Beta 2.5 MeV spiral sector injector of KURRI 150 MeV ADS
FFAG facility, circa 2005 [9]. In both cases, the induction system appears as a large “betatron core”
(red tori, in Ion Beta) which surrounds an arc of the ring

Fig. 5.2 A quite special

betatron: first to be operated,
in 1940 [10, 11], it was also
used for the first
demonstration of a
synchrotron [12, 13]. In
1946, in UK, equipped with
an RF gap, it accelerated
electrons from 4 MeV where
they were brought in
betatron mode, up to 8 MeV
using resonant acceleration
on a fixed orbit

low energy, as the energy gain which the acceleration pulse allows is small (ions at
v << c perform much less revolutions than electrons at .v ≈ c, during the pulse); let
.
alone the large magnet that this would require.
From a historical standpoint the betatron has this particularity that it was used
to demonstrate the concept of longitudinal phase stability in 1946 [12] (preceding
shortly the synchro-cyclotron operation of a 37-in classical cyclotron [14]). This
synchrotron mode of operation of a betatron pioneered the principle of resonant,
cyclic acceleration on a fixed orbit. Within a few years however, the latter took
over for the acceleration of electrons, to potentially far higher energy, with smaller
magnets confined along the electron orbit.
190 5 Betatron

Betatrons when they were first developed and produced were destined for nuclear
physics, as X-ray generators, medical and other applications. Linacs, much more
compact and lighter, easy to include in a gantry system, allowing much higher dose,
have taken over in radiotherapy and in a number of light and compact X-ray source
applications. Betatrons nowadays, in the 3–10 MeV range, are mostly designed for
use as portable, low dose-rate X-ray sources; typical uses include material radiogra-
phy, cargo inspection [15].

5.2 Basic Concepts and Formulæ

Referring to the schematic in Fig. 5.3: the current in the coils is pulsed (an impulse
wave-form at a repetition rate of tens to hundreds Hz) so producing a time-varying
magnetic field . B(t) which induces an accelerating azimuthal electromotive force
along the orbit, while maintaining the electron beam on that constant radius orbit
during acceleration. The vacuum chamber is made from non-conducting material
(glass for instance), or includes non-conducting gaps. Keeping the beam on a closed
orbit around the ring requires the Widerøe condition: the magnetic field along the orbit
must be half the average magnetic field which it circumscribes, a rule established
in 1927 [2]. As in the classical cyclotron (Chap. 3), and in the still to come early
synchrotron (Chap. 8), a tapered gap in the region of the vacuum chamber provides
the necessary weak focusing for stable transverse “betatron oscillations” [16] around
the closed orbit, over the hundreds of thousands of turns that the acceleration to top
energy requires.
The electron energy . E, field . B and radius . R of the fixed orbit are related by
. B R = p/e = β E / ec, where .β → 1 as the electron accelerates. It results that the

maximum energy which can be reached satisfies

. Ê/e = B̂ Rc

Fig. 5.3 Betatron ring,

principle. The equilibrium
orbit is at radius . R where the
field value is one-half of the e−beam
average field encompassed into page
by the orbit. The induction
electric field, resulting from
varying . B(t), is tangent to
the orbit. The magnetic field
in the region of the vacuum
chamber is shaped to provide B(t)
weak focusing R
5.2 Basic Concepts and Formulæ 191

An energy . Ê ≈ 300 MeV can be reached for instance with an orbit radius . R ≈ 1 m
and a guiding field cycled up to . B̂ ≈ 1 T.

5.2.1 Betatron Condition. Acceleration

The inductive electric field .E resulting from the varying magnetic field .B = curlA
satisfies .E = −∂A/∂t. The magnetic vector potential .A has the symmetry of the
current, so . Ar = A y = 0 and .A ≡ As s. Thus the magnetic field components satisfy

∂A 1 ∂r A
. Bs = 0, Br = − , By =
∂y r ∂r

From the expression of . B y it results that the magnetic flux through the area circum-
.
scribed by the orbit of radius . R is
{ R
φ (t) = 2π
. a B y (t) r dr = 2π R A(t) (5.1)
0

Note . Ba ≡ B y = φa /π R 2 the average value of the vertical field component through

the orbital circle. From .Es = −∂As /∂t one gets the accelerating field under the form

R d Ba (t)
. E s (t) = (5.2)
2 dt

With .dp/dt = eE s , and on the other hand .dp/dt = d(eB R)/dt, by integration over
0 → t one gets the betatron condition
.

Ba (t)
. B(t) = (5.3)
2
The guiding field . B(t) along the orbit is one-half of the average field through the area
defined by the orbit. The actual shape of the magnetic field pulse does not matter, as
long as it induces an accelerating emf. Field cycling can for instance take the form

1
. B(t) = B̂ (1 − cos ωt) (5.4)
2

such that .0 ≤ B(t) ≤ B̂. With acceleration taking place over a quarter of a period of
B(t), the duty cycle can get close to 25%. With a pulse over .[−B̂, B̂], two beams can
.
be accelerated both ways, on the positive slope of . B(t) for one, the negative slope
for the other. If they hit a common internal target, two X-ray pulses are emitted, in
opposite directions, at twice the repetition rate .ω/2π .
192 5 Betatron

The gain in energy during acceleration is the same for all electrons, regardless
of their transverse excursions and azimuthal location along the orbit. The beam
spreads around the betatron during the acceleration cycle, as there is no longitudinal
focusing.1

5.2.2 Transverse Motion

The theory of the betatron established the stability of the eponymous oscillations
which electrons undergo under the effect of a radial field index, as they circle in the
vicinity of the closed orbit during acceleration [16]. Elements of theory are introduced
in Chap. 3, Classical Cyclotron, and Chap. 8, Weak Focusing Synchrotron. They can
be referred to for more details.
The properties of betatron oscillations are derived assuming an “adiabatic change
of the magnetic field”, i.e., the magnetic field on the orbit changes slowly compared
to the betatron oscillation. Let’s figure the respective characteristic times of betatron
oscillation and field impulse. The frequencies of the transverse oscillations in a
∂B
structure with cylindrical symmetry and radial field index .0 < n = − ρB00 ∂ xy < 1
write (Eq. 3.18)
√ √
.ω R /ωrev = ν R = 1 − n and ω y /ωrev = ν y = n (5.5)

They are commensurate with the revolution period as, in the betatron,.ν R ≈ ν y ≈ 0.5.
Take for instance . R = 1 m, then .Trev = 2π R/c = 21 ns. On the other hand, assume
60 Hz cycling, meaning magnetic field and momentum ramps of several milliseconds.
This validates the adiabaticity hypothesis, time-varying corrective terms associated
with.B and.p in the equations of transverse motion can be neglected. In this hypothesis
the differential equations of motion (cf. section “Betatron Motion”, Eq. 8.10) have
for solution the harmonic oscillations (cf. section “Betatron Motion”, Eq. 8.21)
⎧ √ ( s )
⎪
⎨ u(s) ≈ βu (s)εu /π cos νu + φ
/ ( s R ) ( s ) (5.6)
.
⎪ εu /π
⎩ u , (s) ≈ − sin νu + φ + αu (s) cos νu + φ
βu (s) R R

with.u(s) standing for.x(s) = δ R(s) (distance to the reference radius. R in the moving
frame), or for . y(s), and other notations as defined in section “Betatron Motion”.
Transverse motion properties include the following [16]:

– on-momentum electrons, whose momentum satisfies . p = eB R, undergo betatron

oscillations around the reference orbit of radius . R;

1 Longitudinal focusing is a property of synchrotron acceleration, see Chap. 7.

5.2 Basic Concepts and Formulæ 193

– electrons with a momentum offset .δp undergo betatron oscillations around a refer-
ence orbit of radius . R + Dδp/ p, with . D = R/(1 − n) the dispersion, a constant
(Eq. 3.20);
√
– the amplitude of these oscillations damps in proportion to .1/ p under the effect
of acceleration, .βγ εu is a constant of the motion (Eq. 8.31).
Beam injection is based on the property that the injected orbit undergoes betatron
oscillations which damp during acceleration, thus allowing part of the injected beam
to miss the injector on successive turns.

5.2.3 Synchroton Radiation

The topic is introduced in this chapter, as the effect of SR on beam dynamics, i.e., orbit
spiraling, was first experienced in a betatron. This theoretical material is resorted to
in the exercises.
Given the energy reached by electrons in a betatron, SR was to be expected.
The emission of radiation by a charged particle had been established about half a
century earlier [18]. SR was resorted to, rightly or not in the early times, in operating
machines as they were reaching meaningful energy, as a possible explanation of some
undesired beam dynamics effects observed. As a matter of fact, the deleterious effect
of SR on the operation of betatrons was pointed out at the time [19]. Measurements
of properties of the radiation were undertaken, that was the beginning of a long story,
still underway...
Some key properties of SR are summarized below, with particular insight in the
stochastic process and the resulting energy loss, as it is the basis of its simulation in
Zgoubi. More is addressed in Chap. 9.

Stochastic Emission of Photons

A detailed theory of SR and its properties can be found in [20]. Energy loss by
synchrotron radiation is comprised of three random processes, namely [21],

– the emission of .k ≥ 0 photons, which abides by the Poisson distribution

/\k −/\
. p(k) = e (5.7)
k!
with
5er0
/\ =< k >=< k 2 >=
. √ Bρ/\θ (5.8)
2h 3
194 5 Betatron

the average number of photons radiated over a trajectory arc./\θ . In this expression,
r = e2 / 4π ε0 m 0 c2 is the classical radius of the electron (.r0 = 2.818 · 10−15 m), .e
. 0
its charge, .m 0 c2 its rest energy, .ε0 = 1 / 36π × 109 , .h = h/2π with .h the Planck
constant;
– the energy .E of the emitted photon(s), which abides by the probability law
{ E/Ec { ∞
3 dE
.P(E/Ec ) = K 5/3 (x)d x (5.9)
5π 0 Ec E/Ec

with . K 5/3 the modified Bessel function and

( )
3hγ 3 c 2ρ
E =
. c or, λc = (5.10)
2ρ 3γ 3

the critical energy (critical wavelength) of the radiation;

– the photon emission angle.ξ with respect to electron momentum vector, a stochastic
quantity which causes scattering of the latter. A cylindrical-symmetric Gaussian
distribution may be assumed for simplicity,

ξ2
. p(ξ ) = exp(− ) (5.11)
2σξ2

For simplicity as well the rms .σξ ≈ 1/γ can be considered independent of photon
energy. The scattering angle of the momentum vector is quite small anyway, and
usually ignored.

Energy Loss

The average energy loss by a ultra-relativistic electron (.β = v/c ≈ 1) over an arc of
trajectory ./\θ in a uniform field . B writes [20]

2 /\θ 2
/\E =
. r0 E 0 γ 4 = r0 ecγ 3 B/\θ (5.12)
3 ρ 3

The stochastic nature of photon emission causes an energy spread which averages to
/ √ √
110 3hc / π ε0 /\θ
σ
. /\E/E = γ 5/2
(5.13)
24E 0 /e ρ
5.3 Exercises 195

Over a revolution (./\θ = 2π in Eq. 5.12), the energy loss writes

E s4 [GeV] 4π r0
Us
. [MeV/turn] = Cγ , Cγ = (5.14)
ρ[m] 3 (m 0 c2 /e)3

For electrons, .Cγ = 8.85852 × 10−5 m/GeV3 .

5.3 Exercises

5.1 Develop a Betatron Magnet in Zgoubi

Solution 5.1
The subroutine which governs the functioning of DIPOLE, dipi.f, can be used as a
template: it can be copy-pasted under a different name, inddip.f for instance, and
modified as needed. The subroutine inddip.f, in addition to computing the guiding
field, will have to allow for a pulsed field, computation of the induction electric field
it entails, and its effect on electron coordinates and momentum.
A keyword such as BETADIPOL can be created (BETATRON is already used! to
simulate a betatron core for slow extraction). A new keyword needs to be added in
LSTKEY.H, and zgoubi.f has to be updated to account for it.
Test this BETADIPOL with the next exercises.

5.2 A 315 MeV Betatron

Solution 5.2
(a) Build an input data file for the Chicago 315 MeV betatron magnet, according to
the parameter list of Table 5.1.
If you did not do Exercise 5.1, then proceed in the following way:
Split DIPOLE into . N = 72 magnetic sectors to simulate the 360.◦ dipole. No
acceleration in this preliminary step.
Check the closed orbit, and the effect of the integration step size.
Check periodic stability and tunes, using TWISS.

Table 5.1 Parameter table (after [17])

Top energy MeV 315
Injection energy keV 135
Orbit radius . R (m) 1.22
Maximum guide field value kG 9.2
Repetition rate Hz 6
. Ḃ T/s To be determined
196 5 Betatron

(b) Simulate an acceleration cycle, assuming a linear . B(t) ramp, for an electron
launched on the closed orbit for simplicity.
Interleave the split DIPOLE with CAVITE[IOPT .= 3] to simulate acceleration.
Note that this will assume longitudinal . E s , with no dependence on transverse coor-
dinates.
Use FAISTORE to store turn-by-turn electron data. Assume no synchrotron radi-
ation in this preliminary step. Check the accelerated orbit.
(c) Simulate the previous acceleration cycle for 2 electrons featuring a paraxial
horizontal excursion for one, a paraxial vertical excursion for the other. Check the
transverse damping of the betatron oscillations.

5.3 Acceleration with Radiation Loss in the 315 MeV Betatron

Solution 5.3
Referring to [17], SR caused a 9% energy loss in the 315 MeV betatron (it was com-
pensated with a voltage impulse from a separate system). In this exercise we check
consistency of the energy loss in the betatron model of Exercise 5.2, with that num-
ber. The SR on/off switch is provided by SRLOSS [22, INDEX]. SRPRNT[PRINT]
can be introduced to log SR data in zgoubi.res and, an effect of the PRINT command,
additional details in zgoubi.SRPRNT.Out.
In order to perform a convergence test on the Monte Carlo SR loss, a 10,000
electron bunch is launched, with 315 MeV energy, for a single turn along the closed
orbit.
DIPOLE field has to be set for 315 MeV; a possibility for that if using data files
from the previous exercise where DIPOLE is set for 135 keV, and update the field
using the global command SCALING.
(a) Get the energy loss per turn .Us , and some of the radiated photon properties
(critical energy, etc.). Check these outcomes against theoretical expectations, sections
“Stochastic Emission of Photons” and “Energy Loss”.
(b) Simulate an acceleration cycle from 0.135 keV to 315 MeV energy region, for
an electron launched on the fixed closed orbit. Check the orbit spiraling. Check the
energy dependence of the synchrotron radiation energy loss. Check the aforemen-
tioned 9% experimental outcome.

5.4 Solutions of Exercises of This Chapter: Betatron

5.1 Develop a Betatron Magnet in Zgoubi
This code development exercise and its benchmarking are left for the reader to
complete.
DIPOLE Fortran source subroutine can be used as a template, namely [pathTo]/
zgoubi-code/zgoubi/dipi.f. This routine includes the assignment of the necessary
dipole parameters for DIPOLE simulation, as specified in zgoubi.dat. It includes
in addition an ENTRY, ‘ENTRY DIPF(...)’, which is called (from chamc.f) during
stepwise raytracing and provides the magnetic field at particle location.
5.4 Solutions of Exercises of This Chapter: Betatron 197

5.2 A 315 MeV Betatron

(a, b) A 315 MeV betatron input data file. Acceleration on the closed orbit.
Acceleration up to 320 MeV is simulated actually, to have some insight a little
beyond 315 MeV. Take kinetic energy (Table 5.1): . E k : 0.135 → 320 MeV, thus . Bρ :
0.00131829454 → 1.069108254786 T m and guiding field . B(t) : 0.0010805693 →
0.876318241628 T (consistent with a maximum 9.2 kG, Table 5.1). Take repetition
rate 6 Hz, assume symmetric saw-tooth like . Ba (t) excitation, hence a ramp up in
.1/12 s. It results . Ḃ = (0.876318241628 − 0.0010805693)/(1/12) = 10.50285 T/s.

{ 5.2 and 5.3 one gets . E s = R Ḃ = 12.81 V/m. Energy gain over a turn
From Eqs.
is .W/e = E s ds = 2π R E s = 2π R 2 Ḃ = 98.211 eV/turn. With a ring comprised of
72 induction modules of 5.◦ angle each, this means.98.211/72=1.364188 eV/module.
Thus use CAVITE[IOPT .= 3,.V̂ = 1.364188] for the induction module.
The number of turns from 0.135 to 320 MeV is .(320 − 0.135) × 106 /98.211 =
3.2565 × 106 . Thus use REBELOTE[NPASS .3.2565 × 106 , IOPT .= 99] for multi-
turn raytracing.

The top file in Table 5.2 defines a 5.◦ induction module. The bottom file in Table 5.2
is set to create a sequence of 72 induction sectors and request acceleration over an
induction cycle.
Tracking outcomes:
• Checking closed orbit and effect of step size.
.

A 1-turn tracking using the input data files of Table 5.2, with acceleration off (tem-
porarily set CAVITE[IOPT.= 0]), results in the following at the bottom of zgoubi.res,
with initial coordinates at OBJET on the left hand side, the final coordinates after a
turn on the right hand side:

The file [b_]zgoubi.fai has the coordinates to greater accuracy, in particular, final
radius after one turn around the ring .Y = 1.2200000000000026 m, and angle .T =
−2.2846308178614549 × 10−12 rad, essentially identical to the starting values
.122 cm and 0 rad respectively. An integration step size near 1 cm ensures closed
orbit closure with such accuracy.
After a .3.25656 × 106 turn acceleration cycle (energy increase shown in Fig. 5.4)
the electron coordinates in zgoubi.res (bottom of the file) appear to be:
198 5 Betatron

Table 5.2 Top input data file: simulation of a 5.◦ “induction sector” module. Bottom input data
file: INCLUDEs this module, and tracks electrons (actually, positrons, here, for convenience) from
135 keV to 320 MeV
5.4 Solutions of Exercises of This Chapter: Betatron 199

Fig. 5.4 Kinetic energy

versus turn in the betatron,
from 0.135 to 320 MeV in
3256566 turns. A graph
obtained using zpop: menu
7; 1/2 to open b_zgoubi.fai;
2/[39,20] for . E versus turn;
7 to plot

These outcomes confirm the stability of the closed orbit, and yields the expected
distance . N × C = 3.25656 × 106 × 2π R = 24.9632 × 106 m, as well as the time
of flight, . N × C/c = 0.0833 s.

• TWISS computation is produced, by running this input data file:

.

The outcome is the following (bottom of zgoubi.res):

200 5 Betatron

Fig. 5.5 Transverse Zgoubi|Zpop Y-1.22, Z (m) vs. Kinetic E (MeV)

0.02 13-10-2023
excursion (markers) and
.(Y0 − 1.22) × p0 / p (solid 0.015
line) versus energy. A graph
obtained using zpop: menu 0.01
7; 1/2 to open [b_]zgoubi.fai;
0.005
2/[20, 2] for .Y versus . E; 7 to
plot; .(Y0 − 1.22) × p0 / p 0.0
graph is superposed using
-.005
option 20
-.01

-.015

-.02
1 2 3 4 5 6 7 8 9

These results show that the closed orbit is at R .= 1.22 m as expected, and that the
lattice is periodically stable.

(c) Transverse damping.

Simulation of the acceleration cycle for 2 electrons featuring a paraxial horizontal
excursion for one, a paraxial vertical excursion for the other uses the bottom file of
Table 5.2 with the following OBJET[KOBJ .= 2, IMAX .= 3]:

Results obtained are displayed in Fig. 5.5. The damping amounts to

/ /
pinitial βγinitial 1
. = =√ = 0.194
pfinal βγfinal 26.46

thus Y and Z motions both damp from .Y0 − 122 = Z 0 = 2 cm at 135 keV to
Ŷ − 122 ≈ Ẑ ≈ 0.39 cm at 320 MeV.
.
A different way to check the convergence is to launch a few particles on an
ellipse, and check that .εY,initial /εY,final = pinitial / pfinal . In that aim, change
OBJET[KOBJ .= 2 ] to the following OBJET[KOBJ .= 8]:
5.4 Solutions of Exercises of This Chapter: Betatron 201

The values of the optical functions in this OBJET are taken from the previous
TWISS.
The input data file has two FAISCEAU (Table 5.2). The first one right after the
object definition by OBJET, the second FAISCEAU at the end of the tracking. In the
resulting listing zgoubi.res the first FAISCEAU provides the following concentration
ellipse data (see Sect. 14.5.1):

which indicate that the initial horizontal ellipse formed by the 30 electrons is cen-
tered at R .= 1.22 m, its surface is twice the concentration ellipse surface, i.e.,
−6
.2 × 5.0000E−07 = 10 m rad, as stated in OBJET. In zgoubi.res the second FAIS-
CEAU (bottom of the file) provides the following concentration ellipse data:

Note that this final concentration ellipse is found offset by .20 µm at 1.22002 m
(from the launch position at R .= 1.22 m), a numerical effect with various possible
causes such as the need for more electrons to better define the positioning of the
matching ellipse, an investigation left to the reader. The surface of the ellipse is
.2 × 1.9065E−08 m rad.

Horizontal phase space ellipses are displayed in Fig. 5.6, showing that
−8 −6
.εY,final /εY,initial = (2 × 1.9066 × 10 )[mm.mrad] /10[mm.mrad] = 0.0381, close to the
expected .1/26.456 = 0.0378. Both are expected much closer actually. Same for
ellipse centering: from the Min-Max Horizontal data at the foot of the graphs,
ellipse center from 30 particles is loosely estimated at .(1.21945 + 1.22171)/2 =
1.22058 m (left), quite different from the statement OBJET[.Y 0 = 122[cm] ], and
at .(1.21980 + 1.22024)/2 = 1.22002 m (right). Zgoubi raytracing accuracy does
allow better precision; this requires more electrons so to properly define the invariant
they lie on, and the parameters of the latter from concentration ellipses.
Checking convergence of the numerical integration for the induction element size
(5.◦ sector or less) and the step size: changing the modular “induction dipole” from a
5.◦ sector to 10 times less, 0.5.◦ , and the step size accordingly, does not change these
results.
202 5 Betatron

Zgoubi|Zpop Zgoubi|Zpop
14-10-2023 Y’ (rad) vs. Y (m) 14-10-2023 Y’ (rad) vs. Y (m)

8E-4

1E-4
4E-4

0.0 0.0

Eps/pi, Beta, Alpha: 4.985E-07 1.1995E+00 -1.3714E-02 Eps/pi, Beta, Alpha: 1.9048E-08 1.3085E+00 1.194E-03

-4E-4
-1E-4

-8E-4
1.22 1.22 1.22 1.221 1.222 1.22 1.22 1.22 1.22

Fig. 5.6 Horizontal phase space positions of a few tens of electrons distributed on an initial ellipse
invariant, and the matching concentration ellipse. Left: 135 keV; right: 320 MeV. The invariant value
is twice the ellipse surface, thus.10−6 and.2 × 1.9066 × 10−8 , respectively. A graph obtained using
zpop: menu 7; 1/2 to open b_zgoubi.fai; 2/[2, 3] for .T versus .Y ; 7 to plot

5.3 Acceleration with Radiation Loss in the 315 MeV Betatron

(a) SR loss over a turn.
The input data file to track 10,000 electrons at 320 MeV, over a turn with SR loss, is
given in Table 5.3
It can be checked in the execution listing zgoubi.res, under DIPOLE data list, that
DIPOLE field is scaled to 320 MeV (from its value set for 135 keV):

Field has been * by scaling factor 798.32335697

SRPRNT delivers statistical data computed from step-by-step Monte Carlo SR.
The results are the following (bottom of zgoubi.res file):

It is an interesting exercise, left to the reader, to compare these step-by-step Monte

Carlo SR outcomes to theoretical expectations. For instance:
5.4 Solutions of Exercises of This Chapter: Betatron 203

Table 5.3 Input data file to track 10,000 electrons over a turn, with SR loss. SCALING[SCL .=
1.3182945462093764 * 798.32335697] is used to scale DIPOLE field (set for 135 keV) to the current
reference rigidity OBJET[BORO .= 1.3182945462093764 * 798.32335697] which corresponds to
a 320 MeV electron

– average energy loss per particle per pass: 0.7252923 keV, with theory’s
0.7148976 keV (Eq. 5.14);
– critical energy of photons (average): 5.7107747E-02 keV, with theory’s
0.0568316 keV (Eq. 5.14).

Note that similar cases of SR simulations may be found in various examples in

zgoubi sourceforge repository, https://sourceforge.net/p/zgoubi/code/HEAD/tree/
branches/exemples/ folder.

(b) An acceleration cycle up to 315 MeV region, including SR.

Building the simulation file starts from Table 5.2. Add SRLOSS as in Table 5.3, and
add REBELOTE[NPASS .= 3256566] for a complete acceleration cycle. Change
OBJET to a single particle. Make sure INCLUDE grabs the “induction” accel-
eration simulation by CAVITE, together with the 5.◦ DIPOLE, i.e., the segment
[#S_Scaling:#E_CAVITE]. The resulting file is as follows:
204 5 Betatron

The simulation is straightforward, refer to the preliminary SR loss simulations above

and to [17] for the interpretation of the tracking outcomes.

References

1. J.P. Slepian, ’X-Ray Tube’, US Patent No 1,645,304, Filed April 1, 1922, published Oct.
11, 1927. See Fig. 2 in F. Scarlat, E. Badita, E. Stancu, A. Scarisoreanu, Basic principles of
conventional and laser driven therapy accelerators. Adv. Med. Imaging Health Inf. 2019(1), 1–
23 (2019). https://kosmospublishers.com/basic-principles-of-conventional-and-laser-driven-
therapy-accelerators/
2. R. Widerøe, A new principle for generation of high voltages. Thesis, Aachen, October 29,
1927. Archiv für Elektrotechnik 21, 387–406 (1928)
3. P.-A. Chamouard, Saturne 2 : 20 years for physics (Sect. 7: The extracted beams of Saturne 2),
in The 20 Years of the Synchrotron Saturne 2, ed. by A. Boudard, P.-A. Chamouard. (World
Scientific, 2000)
4. K. Takayama, KEK digital accelerator and its beam commissioning, in Talk slides, IPAC
2011, September 4–9, 2011, San Sebastian. https://accelconf.web.cern.ch/IPAC2011/talks/
weoba02_talk.pdf
5. S. Nath, Linear induction accelerators at the Los Alamos National Laboratory DARHT facility,
in TH304 Proceedings of Linear Accelerator Conference LINAC2010, Tsukuba, Japan. https://
accelconf.web.cern.ch/LINAC2010/papers/th304.pdf
6. M.A. Green, S. Yu, Superconducting magnets for induction phase-rotation in a Neutrino Fac-
tory. Tech. Note LBNL-48445; SCMAG-749. https://www.osti.gov/servlets/purl/795332
References 205

7. The induction system is apparent in the photos of the MURA accelerators, Figs. 1, 2, 3 and 9, in
K.R. Symon, MURA Days, in Proceedings of the 2003 Particle Accelerator Conference. https://
accelconf.web.cern.ch/p03/PAPERS/WOPA003.PDF Fig. 5.1: Copyrights under license CC-
BY-3.0, https://creativecommons.org/licenses/by/3.0; no change to the material
8. S. Boucher, et al., The Radiatron: a high average current betatron for industrial and security
applications, in TUPP150 Proceedings of EPAC08, Genoa, Italy. https://accelconf.web.cern.
ch/e08/papers/tupp150.pdf
9. K. Okabe, et al., Development of H- injection of proton-FFAG at kurri, in THPEB009 Pro-
ceedings of IPAC’10, Kyoto, Japan. https://accelconf.web.cern.ch/IPAC10/papers/thpeb009.
pdf Fig. 5.1: Copyrights under license CC-BY-3.0, https://creativecommons.org/licenses/by/
3.0; the photo has been trimmed to mostly leave the 2.5 MeV injector
10. A. Sessler, E. Wilson, A Century of Particle Accelerators (World Scientific, 2007)
11. D.W. Kerst, The acceleration of electrons by magnetic induction. Phys. Rev. 60, 47–53 (1941)
12. F.K. Goward, D.E. Barnes, Experimental 8 MeV synchrotron for electron acceleration. Nature
158, 413 (1946)
13. E.J.N. Wilson, Fifty years of synchrotrons, in Proceedings of EPAC96. https://accelconf.web.
cern.ch/e96/PAPERS/ORALS/FRX04A.PDF Fig. 5.2 : Copyrights under license CC-BY-3.0,
https://creativecommons.org/licenses/by/3.0
14. D. Bohm, L. Foldy: Theory of the synchrocyclotron. Phys. Rev. 72, 649–661 (1947). (Demon-
stration of phase stability using Berkeley 37-inch and 184-inch cyclotrons) https://journals.
aps.org/pr/abstract/10.1103/PhysRev.72.649
15. V.A. Fomichev, Mobile accelerator based on ironless pulsed betatron for dynamic objects
radiographing, in IPAC2019, Melbourne, Australia 019-THP. https://accelconf.web.cern.ch/
ipac2019/papers/thpmp026.pdf
16. D.W. Kerst, R. Serber, Electronic orbits in the induction accelerator. Phys. Rev. 60, 53–58
(1941)
17. D.W. Kerst et al., Operation of a 300 MeV betatron. Phys. Rev. 78, 297–1 (1950)
18. A. Liénard, Champ électrique et magnétique produit par une charge concentrée en un point et
animée d’un mouvement quelconque. L’Éclairage Électrique. 16, 5 (1898)
19. D. Iwanenko, I. Pomeranchuk, On the maximal energy attainable in a Betatron. Phys. Rev. 65,
343–1 (1944)
20. A. Hofmann, The Physics of Synchrotron Radiation. Cambridge Monographs on Particle
Physics, Nuclear Physics and Cosmology (20) (Cambridge University Press, 2004)
21. F. Méot, Simulation of radiation damping in rings, using stepwise ray-tracing methods, JINST
10 T06006 (2015). https://doi.org/10.1088/1748-0221/10/06/T06006. http://iopscience.iop.
org/1748-0221/10/06/T06006
22. F. Méot: Zgoubi Users’ Guide. https://www.osti.gov/biblio/1062013-zgoubi-users-guide.
Sourceforge latest version: https://sourceforge.net/p/zgoubi/code/HEAD/tree/trunk/guide/
Zgoubi.pdf

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0
International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing,
adaptation, distribution and reproduction in any medium or format, as long as you give appropriate
credit to the original author(s) and the source, provide a link to the Creative Commons license and
indicate if changes were made.
The images or other third party material in this chapter are included in the chapter’s Creative
Commons license, unless indicated otherwise in a credit line to the material. If material is not
included in the chapter’s Creative Commons license and your intended use is not permitted by
statutory regulation or exceeds the permitted use, you will need to obtain permission directly from
the copyright holder.
Chapter 6
Microtron

Abstract This chapter introduces the microtron, and to the theoretical material
needed for the simulation exercises. It begins with a brief reminder of the historical
context, and continues with the beam optics and acceleration techniques that the
microtron method leans on, relying in that on basic charged particle optics and
acceleration concepts introduced in the previous chapters. It further addresses the
following aspects:

– spiraling accelerated orbits tangenting at the accelerating gap,

– beam recirculation through an accelerating system, via return arcs,
– methods for periodic motion stability,
– harmonic number jump acceleration.
The simulation of a classical microtron only requires one optical element: DIPOLE,
and CAVITE for acceleration. Simulation of a racetrack microtron adds DRIFT to
simulate drift sections, possibly BEND for small steering dipoles along the recir-
culation paths, and a series of CAVITE to simulate a linac section. Particle moni-
toring requires keywords introduced in the previous chapters, such as FAISCEAU,
FAISTORE, PICKUPS, and some others. Beam path and optics optimization use
FIT[2]. SYSTEM is used to, mostly, resort to gnuplot so as to end simulations with
some nice graphs (orbits, fields, or else) obtained by reading data from output files
such as zgoubi.fai (resulting from the use of FAISTORE), zgoubi.plt (resulting from
IL .= 2), or other zgoubi.*.out files resulting from a PRINT command.

Notations Used in the Text

. B magnetic field
. Bρ = p/e magnetic rigidity
.Cn length of .nth orbit
. E; E 0 ; E i ; E n electron energy; at rest; kinetic at injection; on .nth orbit
.e elementary charge
. f rev , . f rf revolution and RF voltage frequencies

© The Author(s) 2024 207

F. Méot, Understanding the Physics of Particle Accelerators, Particle Acceleration
and Detection, https://doi.org/10.1007/978-3-031-59979-8_6
208 6 Microtron

.h harmonic number, an integer, .h = f rf / f rev

.k = BR dd BR radial field index
.l; m; n .(Cn − Cn−1 )/λr f ; .C1 /λr f ; orbit number
.m 0 electron rest mass
.p electron momentum
.R orbit radius
.R F radio-Frequency
.s path variable
. Trev , . Trf revolution and accelerating voltage periods
.v electron velocity
. V (t); V̂ oscillating voltage; its peak value [ ]
, ,
.x , y radial and axial coordinates in the moving frame. (∗), = d(∗)
ds
.β = v/c normalized electron velocity
.γ = E/m 0 c
2
Lorentz relativistic factor
./\E energy gain in accelerating cavity or linac
./\p, .δp momentum offset
.εu Courant–Snyder invariant (.u : x, y)
.λr f RF wavelength

6.1 Introduction

Although a similar geometry and uniform fixed field structure to the classical
cyclotron (Fig. 6.1) and based as well on resonant acceleration using a fixed frequency
oscillating voltage [1–3], it was not until 1944, more than a decade after Lawrence’s
cyclotron, that the original classical microtron concept, a ultra-relativistic electron
beam accelerator, appeared in the literature [1]. A first specimen was built in Canada
four years later [4]. Figure 6.1 shows an early principle schematic of a classical
microtron, with typical parameters as given in Table 6.1.
The concept evolved into the racetrack microtron (RTM), and a 4-sector RTM,
using AVF focusing, was brought to operation a decade later [5]. A technology still
topical today, in specific energy ranges and applications [6]. During this period theo-
retical studies and developments addressed injection efficiency, resonant acceleration
and phase focusing, transverse focusing, modes of operation, etc. [7]. A typical RTM
schematic is shown in Fig. 6.2: return straights ensure recirculation of the bunches
through a linac section, via a pair of 180.◦ dipoles. The latter feature the neces-
sary weak index focusing for vertical stability. Today, MAMI (Mainz Microtron) is
the highest energy microtron installation, a four-RTM cascade delivering 100.µA
CW polarized electron beam up to 1.6 GeV [8]. The RTM is an early stage of the
present-day recirculating linear accelerator (RLA), yet with its recirculating arcs
staked horizontally rather than vertically. In its “double-sided” design (cf. MAMI-C,
43 recirculations in 15 MeV steps) the RTM may be seen as akin to CEBAF-style
two-linac RLA.
6.1 Introduction 209

Fig. 6.1 Schematic of a classical microtron [2]. 1 Vacuum chamber; 2 magnet; 3 accelerating
resonator; 4 waveguide; 5 ferrite; 6 magnetron; 7 electron emitter; 8 high-vacuum pump; 9 extraction
channel

Table 6.1 Typical parameters of an early classical microtron [2]

Energy .10−18 MeV

Magnetic field .1−2 kG

Magnet pole diameter 75 cm

RF 3 GHz
Cavity voltage .≈511 keV

Repetition rate 400 Hz

Average current 50.µA
Bunch emittances, H, V .2 × 1.5, .4 × 15 mm.×mrad

Bunch length .>λrf /20

Microtrons are fixed-field, fixed RF bunch-train accelerators. They provide fast

acceleration, in just a few recirculations through a cavity or a short linac, thus essen-
tially preserving the properties of the bunch out of the source, namely short bunches
210 6 Microtron

Fig. 6.2 A principle schematic of a racetrack microtron [3]. Room is allowed for a high energy
boost linac section

with small transverse emittance and small momentum spread. The isochronism con-
straint requires accelerated bunches to be ultrarelativistic from the first recirculation,
thus the method is of interest for lepton beams. The method is also of interest for
the acceleration of polarized electron beams: an RTM microtron is essentially a long
beam line, with accelerating sections, and so exempt from adverse effects of reso-
nant depolarization as met in cyclic accelerators (see Sects. 4.2.5, 8.2.4, 9.2.7), thus
electron bunch polarization from the source is preserved.
The classical microtron allows a few tens of MeV electron energy range, using an
X-band (9.4 GHz), C-band (5.9 GHz), or S-band (2.8 GHz) high-gradient RF system.
There are various designs of racetrack microtrons, they use short S-band or L-band
(1.3 GHz) standing-wave linacs. Repetition rate is pulsed linacs’ tens of Hz range,
duty cycle up to several %, current up to 100.µA. Racetrack microtrons may use
high fields, up to 2–3 T with normal conducting bends, up to 7–8 T with cryogenic
magnets, this makes them compact electron recirculators [9].
The small 6D emittance electron bunches it delivers make the microtron an
appropriate option in a number of applications, such as injector for synchrotron
light source [9–11], free electron lasers (Fig. 6.3); industrial electron beams; .γ -ray
sources [13, 14]; radiation therapy; photonuclear production of isotopes, etc.
6.2 Basic Concepts and Formulæ 211

Fig. 6.3 Classical microtron injector (white arrow; the black arrow points to its RF cavity mag-
netron), at Kaeri far infrared free electron laser, a 2 m undulator (yellow arrow) down a short beam
line (red arrow) [12]

6.2 Basic Concepts and Formulæ

The basic principles of the microtron are as follows:

• a fixed frequency accelerating RF system,

• bunch acceleration based on harmonic number jump,
• a fixed magnetic field to recirculate the bunches through the RF system, using

– in the case of the classical microtron (Fig. 6.1): a 360.◦ uniform field dipole with
radial index .−1 < k < 0;
– in the case of the RTM (Fig. 6.2):
two or more [5] split sectors, possibly featuring a field index,
return straights to recirculate the beam,
the recirculation path length an integer multiple of the RF wavelength.

6.2.1 Classical Microtron

Referring to Fig. 6.1, and assuming .β ≈ 1, in order to maintain isochronism between

revolution time and oscillating voltage the length of any return orbit is a multiple of
the RF wavelength .λrf . As the energy . E = m 0 c2 is increased .m increases, thus the
revolution frequency decreases, a known property of the classical cyclotron (Chap. 3).
212 6 Microtron

The length of the first orbit is

C1 = mλrf ,
. m ≥ 2, integer (6.1)

with the constraint that it must clear the cavity. In passing, from this constraint an
order of magnitude of the RF can be figured out: assume .C1 ≈ 30 cm, thus . f rf =
c/λrf ≈ 2c/C1 = O(109 ), in the GHz range.
Assume an orbit length increment of .lλrf , .l an integer (note that .l = 1 minimizes
the overall orbit excursion), thus the length of the .nth orbit is

. Cn = [m + (n − 1)l] λrf (6.2)

From . B R = p/e with . p = β E/c, and assuming .β ≈ 1 as a necessary condition for

isochronism of all orbits, one gets the energy increment by the cavity,

ceB c2 eB
/\E = ceB/\R =
. lλrf = /\Trev (6.3)
2π 2π
with
./\Trev = /\C/c = lλrf /c (6.4)

the revolution time increment. The energy of the .nth orbit comes out to be

ceB (m )
. En = [m + (n − 1)l] λrf = /\E +n−1 (6.5)
2π l
Combining the expression for the energy at the first turn

. E 1 = E 0 + E i + /\E (6.6)

with ./\E from Eq. 6.5 yields the energy gain at the accelerating gap,

E0 + Ei
/\E =
. (6.7)
m/l − 1

The dependence of the field strength on .m and .l results, namely

2π /\E 2π E 0 + E i
. B= = (6.8)
ce λrf l ceλrf m − l

The guiding field . B is maximized, and so the magnet size is minimized, for
m − l = 1, which works for .m = 2 (length of first orbit is 2 wavelengths) and .l = 1
.
(orbit length increment is 1 wavelength). With .m > 2 and .l < m, Eq. 6.7 indicates
that the energy increment at the accelerating gap in order to preserve the synchro-
nism is of the order of magnitude of the rest mass of the particle. A constraint which
precludes considering this method for the acceleration of ions.
6.2 Basic Concepts and Formulæ 213

Weak focusing, as in the classical cyclotron (Sect. 3.2.2), ensures transverse

motion stability:
– geometrical focusing maintains horizontal beam confinement in the vicinity of the
closed orbit,
– a small field index (Eq. 3.11) ensures vertical focusing, with minor decrease of
geometrical focusing, and marginal effect on the synchronism and microtron
Eqs. 6.1–6.8.

6.2.2 Racetrack Microtron

In the racetrack microtron the recirculating dipole is split into two halves (Fig. 6.2).
This allows room for a longer linac and greater energy gain, for efficient electron
injection systems, and for beam instrumentation such as orbit correctors and beam
position monitors (BPM). The linac occupies a straight section between the two
magnets, whereas a series of return straight sections connect the trajectory arcs of
increasing radius/energy.
As in the classical microtron, isochronous acceleration requires that the time of
flight of any return orbit be a multiple of the RF period .Trf = 2π/λrf . Thus, as long
as .β ≈ 1, the time difference between two orbits (Eq. 6.4)

. /\Trev = /\C/c = lλrf /c

still holds. With an added drift length . L between the two 180.◦ bends, the energy gain
at the linac writes [15]
E0 + Ei
./\E = (6.9)
m/l − 1 − 2L/lλrf

which reduces to Eq. 6.7 if . L = 0. The dependence of the field strength on m and l
is unchanged compared to the classical microtron, namely (Eq. 6.8),

2π /\E
. B= (6.10)
ce λrf l

The previous two equations show that the bending field and the linac boost must
be adapted to the injection energy in order to satisfy the isochronism condition.
Obviously, the top energy . E n after .n passes in the linac, injection energy . E i and
linac boost ./\E satisfy
. E n = E i + n × /\E (6.11)

with in addition . E i and ./\E linked by Eq. 6.9.

214 6 Microtron

The stable RF phase interval satisfies

2
. tan(/\φ) = (6.12)
πl
A phase slip may result from .β < 1 at the first turn, a few degrees expectedly, and
should be well within the stable phase interval. Additional considerations on energy
variation tolerances, and longitudinal acceptance, can be found for instance in [15]
which addresses the design and properties of a variable energy RTM.
Orbit correction is paramount. It is needed to ensure proper beam steering over
many recirculations, and in particular beam alignment on the linac axis. Techniques
for that include horizontal and vertical steerers placed along the return straights, and
BPMs. Tight orbit control may allow to relax on the dipole field homogeneity and
on their positioning constraint.
Various methods may be resorted to regarding vertical focusing and transverse
stability, depending on the general design of the racetrack. They may include

– wedge focusing [5] (see Sect. 14.4.1),

– active field clamps at the main dipole edges,
– a field index in the main dipoles, as in the classical microtron,
– AVF focusing [16] in the RTM sectors (see Sect. 4.2.1), as in the relativistic
cyclotron,
– additional focusing in the return straights.

6.2.3 Synchrotron Radiation

Synchrotron radiation (SR) matters in high energy microtrons, in a similar way that
it matters in beam lines [17] and recirculation linacs as CEBAF [18].
Effects of SR include energy loss and radial and longitudinal emittance growth.
SR was first observed in a betatron (visually), and its effects on orbit as well, above
300 MeV where it required compensation by an ad hoc RF system. For this reason it
is introduced in the Betatron chapter (Chap. 5), which can be referred to. The orbit
spiraling which energy loss by SR causes in a high .γ RTM (cf. section “Energy
Loss”, Eq. 5.14) requires compensation measures.
Emittance growth upon SR matters in high.γ rings, whose future possibly includes
the muon collider [19] and other FCC lepton and hadron collider rings [20]. It is
introduced in that context, in the Strong Focusing Synchrotron chapter, Sect. 9.2.5,
which can be referred to. As a matter of fact, in a GeV range microtron SR induced
emittance growth matters as well, in relation with linac pipe aperture for instance.
6.3 Exercises 215

6.3 Exercises

Note: Some of the input data files for these simulations are available in zgoubi
sourceforge repository at

.[pathTo]/branches/exemples/book/zgoubiMaterial/microtron/

6.1 Build a Classical Microtron

Solution 6.1
A 10-pass classical microtron simulation is worked out in this exercise, using S-band
RF. Machine parameters are checked first, by raytracing. A bunch is then tracked,
over an acceleration cycle.
(a) Build an input data file for a classical microtron (following the principle sketch
of Fig. 6.1) with parameters taken from Table 6.2 (after Ref. [14]). DIPOLE can be
used for the guiding field simulation (exercises in cyclotron Chaps. 3 and 4 can be
referred to, field simulations are similar). Be sure to introduce a weak field index
in order to ensure vertical motion stability. CAVITE[IOPT .= 3] can be used as a
zero-length accelerating system.
Validate the input data file by tracking an on-momentum electron, on the reference
orbit, from injection to extraction energy. From the raytracing outcomes, check

– the expected circumference of the .nth orbit, .Cn (Eq. 6.2),

– the energy . E n of the .nth orbit (Eq. 6.5).

(b) Track a.2 × 103 -electron bunch over a complete acceleration cycle. Take initial
bunch transverse emittances .εx = ε y = 0.1 π μm, and momentum spread .δp/ p =
10−3 .
Produce a graph of a few trajectories in the laboratory frame.
Produce graphs of the final transverse and longitudinal phase spaces.

Table 6.2 Parameters of a 9.5 MeV classical microtron, after Ref. [14]. Injection energy . E i is at
cavity entrance
Injection energy (. E i ) 0.409 MeV
Extraction energy (. E x ) 9.6 MeV
Number of orbits (.n) 10
Energy gain in accel. gap (./\E) To be determined
RF (. f rf ) 2.8 GHz
Length of first orbit (.C1 ) .2 × λrf

Length of .nth orbit (.Cn ) .(n + 1)λrf

216 6 Microtron

6.2 Build a 100 MeV Racetrack Microtron

Solution 6.2
The microtron schematic of Fig. 6.2 is considered in this exercise. Take the following
parameters [23, Sect. 3.1]: 5 MeV linac, 3 GHz RF, injection energy 50 keV, spacing
between the half-dipoles 1 m, final energy 100 MeV in 20 linac passes.
DIPOLE can be used for the sectors of the 180.◦ bends. Assume hard-edge for
simplicity. BEND is convenient to simulate correction dipoles in the return straights,
one at each end for instance, if necessary. DRIFT is used for the field-free spaces. A
series of 6 CAVITE elements can be used to simulate the linac section.
(a) Assemble the RTM: the two double-sector 180.◦ bends, drift sections with
correction dipoles, linac section. Check the geometry by producing a graph of the
accelerated orbit. No need to set up all 20 passes, 3–4 passes are enough to establish
the effectiveness of the simulation and assess the various parameter adjustments to
be performed to make it work.
(b) Check the transverse stability of the optics.
(c) Accelerate a 6D bunch. Check the evolution of its parameters: energy, trans-
verse and longitudinal phase spaces.

6.4 Solutions of Exercises of This Chapter: Microtron

6.1 Build a Classical Microtron

(a) A classical microtron, input data file.
Based on the microtron parameters given in Table 6.2, the simulation input file of
Table 6.3 results. In particular,

– a single particle is launched under OBJET; its launch radius is taken in (Fig. 6.4)

.0 < Y0 < C1 /π = 2λrf /π ⇒ 0 < Y0 < 6.8 cm

a necessary condition for DIPOLE to function, as per its geometrical definition [21,
Fig. 9];
– the integration step size in DIPOLE is taken substantially less than .C1 = 2λrf =
21.4 cm, to ensure accurate numerical integration of the equations of motion;
– the energy gain under CAVITE is (Eq. 6.7 with .l = m − l = 1) ./\E = E 0 + E i <
2E i ;
– REBELOTE[IPASS .= 9] ensures a 10-turn acceleration cycle;
– in addition: a weak filed index is added (Eq. 3.11), to ensure vertical motion stabil-
ity. The value is taken from the Classical Cyclotron Chap. 3, Exercise 3.6, namely,
|
R0 ∂ B y ||
k= = −0.03
B0 ∂ R | R=R0 ,y=0
.
6.4 Solutions of Exercises of This Chapter: Microtron 217

Table 6.3 Simulation input data file microtron_one360degDipole.dat. An analytical modeling of

a dipole magnet field, using DIPOLE. That file defines the labels (LABEL1 type [21, Sect. 7.7])
#S_piMicrotronSector and #E_piMicrotronSector for INCLUDEs in subsequent questions. It also
accelerates a single particle launched in the vicinity of the reference orbit
218 6 Microtron

Fig. 6.4 Accelerated orbit

from 0.4 to 9.5 MeV, in a
microtron operated classical
cyclotron magnet

This is small enough a value that the general microtron Eqs. 6.1–6.8 still hold,
with marginal perturbation. Note that .k is defined with respect to the center of the
dipole, as a consequence the magnetic field over a revolution is not homogeneous;
– a marginal vertical motion is added (. Z 0 = 0.1 mm), for monitoring the vertical
focusing. It is small enough that the trajectory of the particle only marginally
departs from the reference accelerated orbit in the median plane. That trajectory
can thus still be used to check raytracing outcomes against theory, Eqs. 6.1–6.8.
The input data file (Table 6.3) is set for a 10-turn tracking of a single particle
launched near the reference orbit. The resulting accelerated orbit, from injection to
extraction energy, is displayed in Fig. 6.4. The dependence of orbit circumference and
energy on turn number is displayed in Fig. 6.5, with comparison to theory. Figures 6.6
and 6.7 show the field experienced along, and coordinates of the accelerated orbit.

(b) Acceleration of a 6D bunch.

The simulation file is given in Table 6.4. One or the other of the available variables,
or several: RF system parameters (synchronous phase and voltage under CAVITE);
the initial bunch conditions (injection phase via coordinate .s0 under MCOBJET,
injection energy); the dipole field, need adjustment to maximize the transmission.
This optimization exercise is left to the reader. A possibility is to develop, or use
existing algorithms, for beam steering and for the longitudinal motion. A brute force
method can be used as well, consisting in scanning the variables, with the objective of
maximizing the number of particles transmitted, with proper final energy. Zgoubi’s
FIT procedure can also resorted to, it allows such constraints as the average value of
one or the other of the final coordinates of the particles (FIT[IC .= 3]) [21], including
their momentum. More constraints, adapted to this particular type of problem, can
be added in the zgoubi source subroutine concerned, ff.f [22].
6.4 Solutions of Exercises of This Chapter: Microtron 219

Fig. 6.5 Dependence of orbit circumference (square markers) and energy (circles) on turn number,
from raytracing, data read from zgoubi.fai [21, Sect. 8.2]. Solid lines: theory, Eqs. 6.2 and 6.5,
respectively. Note: orbit length comes closer to theory if the field index .k is set to zero

Fig. 6.6 The magnetic field

varies along the accelerated
orbit under the effect of the
transverse field index,
however only weakly, so
preserving near-synchronism
with the RF voltage
220 6 Microtron

Fig. 6.7 Horizontal (red, fast oscillation) and vertical (blue, slow oscillation) coordinates of the
accelerated particle, over the 10 revolutions

Table 6.4 Simulation input data file for 6D bunch acceleration. CAVITE[IOPT .= 2] is used to
accelerate, accounting for RF phase motion
References 221

6.2 Build a 100 MeV Racetrack Microtron

Setting up this simulation, and checking it, is left to the reader. The input data file of
the previous exercise, Table 6.3, can be used as a starting point for a split DIPOLE
in the recirculation path.
Guidance in setting up the optical parameters for this RTM can be found in [23,
Sect. 3.1].

It is a good idea for such recirculator simulation from start to end in a single job, to
resort to GOTO keyword. GOTO in this context is used to switch the beam to proper
sub-systems (i.e., return straights, 180.◦ dipoles, linac). In this case it may be required
to end the job using FINISH (END may not work in this context). AUTOREF also
is useful to recenter the beam when presented at these sub-systems.
Guidance in setting up the input data file using these keywords can be found in
an existing complete recirculating linac simulation, an energy recovery linac (ERL),
12 passes up 11 passes down, in https://sourceforge.net/p/zgoubi/code/HEAD/tree/
branches/exemples/didacticExercises/LR-eRHIC/ folder. This ERL simulation is
based on the keywords
– REBELOTE[IOPT .= 1] in combination with GOTO[PASS#] and GOTO
[GOBACK] to switch the beam to proper subsystems (return loops, spreader,
combiner and linac),
– REBELOTE[IOPT .= 1, LMNT .= DRIFT[DltaPhase]] to flip the RF phase by pi
at entrance to the linac, for energy recovery from pass number 14 on,
– abundant use of INCLUDE so to allow separate data files for the recirculating
channel(s), spreaders, combiners and linac–which has the merit additional of sim-
plifying the main input data file,
– FINISH, which ends the job, in lieu of END which cannot be used in this context.

References

1. V.I. Veksler: Proc. USSR Acad. Sci. 43, 346 (1944); J. Phys. USSR 9, 153 (1945)
2. S.P. Kapitsa, The microtron and areas of its application. Translated from Atomnaya Énergiya
18(3), 203–209 (1965)
3. P. Lidbjörk, The microtron, in CERN Accelerator School, Fifth General Accelerator
Physics Course, Proceedings, University of Jyväskylä, Finland, 7–18 September 1992,
vol. 2. Report number: CERN-94-01, CERN-YELLOW-94-01. http://cds.cern.ch/record/
235242/files/CERN-94-01-V2.pdf. Figure 7.18: Copyrights under license CC-BY-3.0, https://
creativecommons.org/licenses/by/3.0; no change to the material
4. W.J. Henderson, H. Le Caine, R. Montalbetti, A magnetic resonance accelerator for electrons.
Nature 162, 699–700 (1948). https://doi.org/10.1038/162699a0
5. E. Brannen, H. Froelich, Preliminary operation of a four-sector racetrack microtron. J. App.
Phys. 32, 1179 (1961)
6. C. Hori, et al., Optical design of AVF weak-focusing accelerator, in TUP036, Proceedings of
the 22nd International Conference on Cyclotrons and Their Applications, Cape Town, South
Africa (2019), pp. 242–244. https://accelconf.web.cern.ch/cyclotrons2019/papers/tup036.pdf
222 6 Microtron

7. A.P. Grinberg, The microtron. Soviet Physics Uspekhi 4(6), 857–879 (1962). https://doi.org/
10.1070/PU1962v004n06ABEH003391
8. M. Dehn, K. Aulenbacher, R. Heine, et al., The MAMI-C accelerator. Eur. Phys. J. Spec. Top.
198, 19 (2011). https://doi.org/10.1140/epjst/e2011-01481-4. https://www.blogs.uni-mainz.
de/fb08-nuclear-physics/accelerators-mami-mesa/the-mainz-microtron/
9. T. Hori, Ten years of compact synchrotron light source AURORA, in WEP55, Proceedings
of the 1999 Particle Accelerator Conference, New York (1999). https://accelconf.web.cern.ch/
p99/PAPERS/WEP55.PDF
10. Nadji, et al., Status of SESAME project. Proceedings of PAC09, Vancouver, BC, Canada
WE5RFP022. https://accelconf.web.cern.ch/PAC2009/papers/we5rfp022.pdf
11. W.H.C. Theuws, et al., The 75 MeV racetrack microtron Eindhoven, in Proceedings of the
Linac 96 Conference. https://accelconf.web.cern.ch/l96/PAPERS/MOP18.PDF
12. Y.U. Jeong, Compact terahertz free-electron laser as a users facility, in Proceedings of
APAC 2004, Gyeongju, Korea. Fig. 6.3: Copyrights under license CC-BY-3.0, https://
creativecommons.org/licenses/by/3.0; no change to the material
13. R. Hajima, et al., Compact gamma-ray source for non-destructive detection of nuclear material
in cargo. THPS098, in Proceedings of IPAC2011, San Sebastián, Spain. https://accelconf.web.
cern.ch/IPAC2011/papers/thps098.pdf
14. R.J. Abrams, et al., Compact, microtron-based gamma source, in THPMR052, Proceedings of
IPAC2016, Busan, Korea. https://accelconf.web.cern.ch/ipac2016/papers/thpmr052.pdf
15. W.H.C. Theuws, et al., Continuous electron-energy variation of the Eindhoven racetrack
microtron, in 17th IEEE Particle Accelerator Conference (PAC 97). https://accelconf.web.
cern.ch/pac97/papers/pdf/7W020.PDF
16. H.R. Frœlich, J.J. Manca, Performance of a multicavity racetrack microtron. IEEE Trans.
Nucl. Sci. (Proceedings of PAC75) NS-22(3) (1975). https://accelconf.web.cern.ch/p75/PDF/
PAC1975_1758.PDF
17. G. Leleux, et al., Synhrotron radiation perturbations in long beam lines, in Proceeding of the
PAC 1991 Accelerator Conference, May 6–9, 1991 San Francisco, California, USA. https://
accelconf.web.cern.ch/p91/PDF/PAC1991_0517.PDF
18. D.R. Douglas, et al., Control of synchrotron radiation effects during recirculation, in IPAC2015,
Richmond, VA, USA. https://accelconf.web.cern.ch/IPAC2015/papers/tupma035.pdf
19. B.J. King, Further studies on the prospects for many-TeV muon colliders, inProceedings of the
PAC 2001 Accelerator Conference, 18–22 Jun 2001, Chicago, IL, USA. https://accelconf.web.
cern.ch/p01/PAPERS/RPPH314.PDF
20. M. Benedikt, F. Zimmermann, Status of the future circular collider study, in Proceed-
ings of RuPAC2016, St. Petersburg, Russia. https://accelconf.web.cern.ch/rupac2016/papers/
tuymh01.pdf
21. F. Méot, Zgoubi users’ guide. https://www.osti.gov/biblio/1062013-zgoubi-users-guide.
Sourceforge latest version: https://sourceforge.net/p/zgoubi/code/HEAD/tree/trunk/guide/
Zgoubi.pdf
22. https://sourceforge.net/p/zgoubi/code/HEAD/tree/trunk/zgoubi/ff.f
23. P. Lidbjörk, The microtron, in CERN Accelerator School, Fifth General Accelerator Physics
Course, Proceedings, University of Jyväskylä, Finland, 7–18 September 1992, vol. 2.
Report number: CERN-94-01, CERN-YELLOW-94-01. http://cds.cern.ch/record/235242/
files/CERN-94-01-V2.pdf
References 223

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0
International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing,
adaptation, distribution and reproduction in any medium or format, as long as you give appropriate
credit to the original author(s) and the source, provide a link to the Creative Commons license and
indicate if changes were made.
The images or other third party material in this chapter are included in the chapter’s Creative
Commons license, unless indicated otherwise in a credit line to the material. If material is not
included in the chapter’s Creative Commons license and your intended use is not permitted by
statutory regulation or exceeds the permitted use, you will need to obtain permission directly from
the copyright holder.
Chapter 7
Synchrocyclotron

Abstract This chapter introduces the concept of phase focusing by synchronous

acceleration, and the synchrocyclotron which confirmed the principle. Synchrocy-
clotron style of acceleration in a fixed field alternating gradient accelerator (FFAG) is
also addressed. The theoretical material needed for the simulation exercises is essen-
tially that of the Weak Focusing Synchrotron, Chap. 8, regarding phase stability, and
that of the Classical Cyclotron, Chap. 3, or FFAG optics, Chap. 10, regarding trans-
verse stability. The chapter begins with a brief reminder of the historical context, and
continues with the theoretical material which the synchrocyclotron optics and accel-
eration techniques lean on. The simulation of a synchrocyclotron is achieved using
just three keywords: DIPOLE for the magnet, and CAVITE and SCALING for accel-
eration. FFAG dipoles have their specific keywords, FFAG and FFAG-SPI (Chap. 10).
Particle monitoring uses FAISCEAU, FAISTORE, and some others. Optics match-
ing and optimization, and the design of RF programs as well, use FIT[2]. INCLUDE
is resorted to, although there is no obligation, in order mostly to simplify the input
data files. SYSTEM calls to gnuplot scripts allow ending simulations with various
graphs; gnuplot reads data from output files such as zgoubi.fai (produced by FAI-
STORE), zgoubi.plt (resulting from IL .= 2) and from files zgoubi.*.out resulting
from a PRINT command.

Notations Used in the Text

. B magnetic field value

. Bρ = p/q; Bρ0 particle rigidity; reference rigidity
.C; .C0 orbit length
. E; E s particle energy, . E = γ m 0 c2 ; synchronous energy
. f rev , . f rf = h f rev revolution and RF voltage frequencies
.h RF harmonic number, .h = f rf / f rev
.m; .m 0 ; . M particle mass; rest mass; mass in units of MeV/c.2
.k =
R dB
B dR
radial field index
.p; . p; . p0 momentum vector; its modulus; reference
.q particle charge

© The Author(s) 2024 225

F. Méot, Understanding the Physics of Particle Accelerators, Particle Acceleration
and Detection, https://doi.org/10.1007/978-3-031-59979-8_7
226 7 Synchrocyclotron

.R average orbit radius, .C = 2π R

.R, θ, y particle coordinates, radial, azimuthal, axial
.α momentum compaction, or trajectory deviation
.β = v/c; .β0 ; .βs normalized particle velocity; reference; synchronous
.γ = E/m 0 c ; .γtr Lorentz relativistic factor; transition .γ
2

.νR, θ wave numbers, radial and axial

.φ; φs particle phase at voltage gap; synchronous phase

7.1 Introduction

The synchrocyclotron (SC) accelerator is an outcome of the 1945 concept of phase

focusing synchronous acceleration [1–3]. Demonstration of the latter successfully
used a small classical cyclotron [4] (following closely a proof-of-principle in fixed
closed orbit regime, using a betatron [5]).
Synchronous acceleration opened the way to the highest energies: this is the accel-
eration method in today’s high energy colliders. Acceleration techniques at the time
had intrinsic energy limitations: electrostatic generators around a few MeV due to
insulation break-down at high electric field; the classical cyclotron, in the few MeV
ion energy range (a few keV in the case of electrons) due to the loss of isochro-
nism resulting from relativistic increase in mass (Chap. 3); the bulky isochronous
ion cyclotron in the GeV range in relation with extraction efficiency (Chap. 4); the
betatron due to the loss of the Widerøe condition resulting from synchrotron radiation
(Chap. 5).
Phase focusing in a SC requires varying the frequency of the accelerating voltage,
for it to follow the increase of the revolution period (Eq. 3.3) as the accelerated
bunch spirals outward. A consequence is that the acceleration has to be cycled, at
a rate determined by the time it takes to bring a bunch from injection to extraction
energy. The repetition rate of the RF frequency cycling is typically of the order of
.10 − 10 Hz, determined by orbit size, energy, and RF voltage. This is orders of
2 3

magnitude below cyclotron CW regime (with a bunch repetition rate of typically

tens of MHz) as is the delivered average current.
The aforementioned successful demonstration using a 37-inch cyclotron resulted
in the modification of Berkeley 184-inch cyclotron (Fig. 3.3), under commissioning
at the time, into a synchrocyclotron which was brought into operation in 1946, and
allowed producing 200 MeV proton beams and 400 MeV alpha-particle beams.
The highest beam energy from a frequency modulated cyclotron was achieved in
Gatchina (Leningrad, Russia) in the late 1960s (Fig. 7.1), producing 1 GeV proton
bunches at a 40–60 Hz repetition rate. A 600 MeV SC was CERN’s first accelerator,
providing beams from 1957, for particle and nuclear physics; only leaving particle
physics to the CERN PS in 1964; supplying short-lived beam to ISOLDE from 1967
until its shut down in 1990. Its parameters, typical of a high energy SC, are given in
Table 7.1; note the small peak accelerating voltage, more than an order of magnitude
what an isochronous cyclotron requires. The SC technology is still topical today,
7.1 Introduction 227

Fig. 7.1 Gatchina 1 GeV synchrocyclotron [6]

Table 7.1 Parameters of the CERN 600 MeV SC

Magnet pole diameter 5m
Magnetic field 1.9 T
Peak voltage 25 kV
RF sweep 29.→ 16.5 MHz
Repetition rate 50 Hz
Average current 1.µA

in particular in protontherapy application as the use of superconducting magnet

technology allows for compact devices (Fig. 7.2).
A typical history line in that respect is Orsay synchrocyclotron (Fig. 7.3, param-
eters in Table 7.2): since 1991 it was one of the two protontherapy accelerators in
hospital environment in France [9] (with MEDICYC in Nice [10]). The 157 MeV,
450 Hz repetition rate SC delivered a first beam in 1957, a typical nuclear physics
research installation; the facility was shut-down in 1975 for evolution to 200 MeV;
in 1993 the installation was converted to a hadrontherapy hospital, IC-CPO (Institut
Curie-Centre de Protontherapie d’Orsay).
In a general manner the synchrocyclotron method can be understood as applying
the phase focusing technique in a fixed-field ring accelerator. This is one way FFAGs
are operated (they may also use induction acceleration [11, 12]—Chap. 10). By
contrast with the classical cyclotron, FFAGs use high gradient radial or spiral sector
dipoles and feature strong focusing optics (Chap. 10).
228 7 Synchrocyclotron

Fig. 7.2 The 230 MeV superconducting S2C2 [7], a 1 kHz repetition rate compact synchrocyclotron
for protontherapy. Parameters: RF frequency 60–90 MHz; magnetic field 5 T; overall diameter 2.5 m;
weight 50 Ton

Fig. 7.3 Layout of Orsay 200 MeV synchrocyclotron [8], a 450 Hz repetition rate machine using a
rotating condenser for RF cycling

Table 7.2 Parameters of the Magnet pole diameter 2.4 m

Orsay 200 MeV SC, first a
nuclear physics instrument, Magnetic field 1.6 T
then converted to Peak voltage 25 kV
hadrontherapy RF sweep 25.→ 20 MHz
Repetition rate 450 Hz
Average current 3.µA
7.2 Basic Concepts and Formulæ 229

7.2 Basic Concepts and Formulæ

The classical cyclotron offered the opportunity of implementing the concept of phase
stability, using existing technology. This further allowed a leap in ion energy, up to
GeV energy range [6].
In the small classical cyclotron used for the demonstration, the oscillating electric
voltage was applied between a dee and a flat electrode facing it. The voltage can be
low, in the kVolt range, an easier technology compared to hundreds of kVolts required
by the isochronism condition in a cyclotron. Many more turns are thus needed, of
the order of .105 compared to a few 100 in a cyclotron, however a large number of
turns no longer matters thanks to the phase focusing.
A drawback of synchronous acceleration in a cyclotron is that the RF system,
thus bunch delivery, has to be cycled, due to the time of flight variation (increasing
with energy). Only particles which maintain correct RF phase at the accelerating
gaps, within a few degrees, are held in a bunch. The magnetic field is fixed, though
(by contrast with pulsed synchrotrons which also require cycling the magnets) thus
allowing a large repetition rate, nevertheless .4−5 orders of magnitudes below a
cyclotron CW regime, to the detriment of the average current.
In FFAGs a synchronous RF system is comprised of modular cavities, providing
one or more accelerating gaps, similar to the RF technology found in synchrotrons.
Drift sections between the dipoles provide the space for inserting these cavities
(Chap. 10).

7.2.1 Phase Stability

The two accelerator lattice species, weak√and strong focusing, differ by the energy
range of their transition gamma, .γtr = 1/ α (.γtr is defined in Sect. 8.2.2), a property
of the lattice which determines two different phase focusing and thus acceleration
regimes: either below transition, or above transition.
Weak focusing results in .γtr ≈ ν R (Eq. 8.34), while due to revolution symme-
try .ν R < 1 (Eq. 3.19). Thus in a classical cyclotron .γtr < 1 < γ , regardless of .γ ,
acceleration is above transition always, which in a practical manner means on the
negative slope of the accelerating wave. This is sketched in Fig. 7.4: a particle with
slightly greater energy than the synchronous particle takes more time to go around
the ring (Eq. 3.3) (path length increase is larger than velocity increase), it tends to
arrive later at the RF gap (at .φ > φs ), thus experiences smaller voltage which tends
to speed it up. A particle with a lower energy is faster and arrives at the gap earlier,
.φ < φs , it experiences greater voltage which tends to slow it down. In both cases the
non-synchronous particle is pulled towards the synchronous phase, this results in an
overall stable oscillatory motion around .φs , the particles stay bunched in the vicinity
of the synchronous phase.
230 7 Synchrocyclotron

Fig. 7.4 A sketch of the mechanism of phase stability in a synchrocyclotron. Stability occurs for
particles falling in the vicinity of a constant synchronous phase .φs , turn after turn, at locations B,
B’, B” ...

√
In an FFAG .γtr ≈ ν R ≈ 1 + k > 1 (cf. Sect. 10.2, Eqs. 10.13 and 10.18), with
.k the field index. Generally .γ < γtr : particle acceleration is below transition (cf.

Sect. 8.2.2, Fig. 8.15).

The phase focusing technique results in synchrotron oscillations, the particle
motion in the longitudinal (momentum, RF phase) phase space. This is addressed in
Sect. 8.2.2, Weak Focusing Synchrotron chapter.

7.2.2 Transverse Stability

The classical cyclotron features weak vertical focusing based on a small.d B/d R < 0,
and geometrical horizontal focusing. This ensures periodic stability, the technique is
addressed in Sect. 3.2.2.
FFAGs instead are based on strong focusing optics, using high transverse gradient
combined function dipoles (typically, a dB/dR gradient of several T/m, as in strong
focusing synchrotrons—Sect 9.2.1). FFAG lattices use radial sector magnets, yield-
ing “Alternating Gradient” optics, the method is addressed in Sect. 10.2. They also
use spiral sector dipoles, the method is addressed in Sects. 4.2.1 and 10.2.2.
JINR phasotron, a 680 MeV, .≈ 3 m radius, proton synchrocyclotron has a similar
structure, of radially increasing average magnetic field, and spiral azimuthal field
variation for periodic stability [13].
7.4 Solutions of Exercises of This Chapter: Synchrocyclotron 231

7.3 Exercises

7.1 Operate a Cyclotron Dipole in Synchrocyclotron Mode

Solution 7.1
Using a dipole magnet simulation taken from the classical cyclotron exercises
(Sect. 3.3), add SCALING[IOPT .= −2] and CAVITE[IOPT .= 6] to simulate an
acceleration cycle, in the synchrocyclotron mode.

7.2 Operate an FFAG in Synchrocyclotron Mode

Solution 7.2
Using a lattice simulation taken from the scaling FFAG exercises (radial, Sect. 10.3.1,
or spiral, Sect.10.3.2), add SCALING[IOPT .= −2] and CAVITE[IOPT .= 6] to sim-
ulate an acceleration cycle, in the synchrocyclotron mode.

7.4 Solutions of Exercises of This Chapter:

Synchrocyclotron

7.1 Operate a Cyclotron Dipole in Synchrocyclotron Mode

The problem requires a classical cyclotron dipole model. This can be based, indif-
ferently, on
– TOSCA keyword [14] and a field map: the method is devised in Exercise 3.6,
with solution 3.6. Zgoubi input data file given in Table 3.16, “FieldMapSec-
torIndex.inc”, which resorts to a field map of a 180.◦ sector dipole with index, can
be used;
– or DIPOLE keyword [14] for an analytical field model: the method is devised in
Exercise 3.6 as well (3.6 (c)). The simulation of a sector DIPOLE with index given
in Table 3.18, “sectorWithIndex.inc”, can replace TOSCA in the input data file of
Table 3.16.

Acceleration
Synchrotron acceleration then needs to be installed. CAVITE[IOPT .= 6] is used,
the option IOPT .= 6 allows reading the RF voltage law (frequency, voltage, etc.)
from a ancillary file zgoubi.freqLaw.In, see below. A similar problem is solved in
the case of synchrocyclotron operation of a radial FFAG lattice and can be referred
to, Exercise 10.5. The same procedure is repeated here, it comprises three steps, as
follows.
(i) Find a set of closed orbits in the acceleration range of concern, 20 keV to 6 MeV
about, and their revolution period (a few tens of closed orbits is fine, it does not need
to be turn-by-turn, zgoubi will interpolate from zgoubi.freqLaw.In content). This
can be performed using FIT, the input data file for that is given in Table 7.3, orbits
232 7 Synchrocyclotron

Table 7.3 Simulation input data file orbit_20to6000keV_FIT.dat. It finds 30 cyclotron orbits,
evenly spaced from 200 keV to 6 MeV, using REBELOTE and FIT. This file also defines the
LABEL1s #S_halfDipole_SC and #S_halfDipole_SC for use in subsequent data files

and time of flights produced by its execution are displayed in Fig. 7.5. The output
file of interest here is orbits.fai, it is needed in step (ii).
(ii) Run an interface program, essentially a read-write procedure: read from
orbits.fai, write in zgoubi.freqLaw.In with the proper formatting (Table 7.4).
(iii) Build the appropriate zgoubi input data file for acceleration (Table 7.5).
This requires the following.
Either one of the aforementioned TOSCA modeling or DIPOLE modeling files
can be used as a starting point, mutatis mutandis, as follows:

– PARTICUL[PROTON] is necessary as CAVITE is used: it allows converting

energy change in rigidity change (zgoubi pushes particles using rigidity),
7.4 Solutions of Exercises of This Chapter: Synchrocyclotron 233

Fig. 7.5 Left: thirty constant-energy closed orbits across a half-dipole, obtained by running the
input data file of Table 7.3. Stepwise integration data are read from zgoubi.plt, filled up upon
IL .= 2 under TOSCA. Right, right vertical axis (circles): time of flight along these half-circle
orbits, versus rigidity; left vertical axis (squares): orbital radius

Table 7.4 The content of zgoubi.freqLaw.In (top and bottom parts), as read by zgoubi when using
CAVITE[IOPT .= 6]. Zgoubi actually only uses the turn number, column 1 (and will interpolate
as needed), and the revolution time which is the cumulated time-of-flight across the cells, column 4

– SCALING[IOPT .= −2] provides the RF program to CAVITE (by reading it from

an external file, zgoubi.freqLaw.In). The RF program here simply provides the
turn dependence of revolution time (Table 7.4),
– CAVITE[IOPT .= 6] boosts the particle(s) at each pass, following that pre-defined
RF program,
– REBELOTE sends the execution pointer back to the top of the input data file,
for multiturn tracking. A .2 × 20 kV acceleration rate per turn may be obtained
from peak voltage .V̂ = 40 kV and synchronous phase .φs = 30◦ at the accelerating
gap. This determines the number of passes for a 0.02.→6 MeV cycle, namely,
.(6 − 0.02)/0.02 = 300, or 150 turns in the dipole,
– FAISTORE stores turn-by-turn particle data (to some user defined file, e.g.,
zgoubi.fai).

Outcomes of these synchrocyclotron acceleration simulations are displayed in

Fig. 7.6.
234 7 Synchrocyclotron

Table 7.5 Simulation input data file for synchronous acceleration in a cyclotron dipole, from 20 keV
to .∼ 6 MeV, in 150 turns.

Fig. 7.6 Left: spiral trajectory of the synchronous particle over 150 turns from 0.02 to 6 MeV
about, obtained by running the input data file of Table 7.5. Stepwise integration data are read from
zgoubi.plt, filled up upon IL .= 2 under TOSCA. Right, right vertical axis (circles): increasing
energy, half-turn by half-turn; left vertical axis (squares): increasing orbital radius
References 235

7.2 Operate an FFAG in Synchrocyclotron Mode

Synchronous acceleration is generally used in FFAGs—in addition to induction accel-
eration, for instance in the MURA FFAGs [11] and in the ion-Beta at the KURRI
Institute [12].
The present problem is treated as part of the FFAG Chapter exercises, Sects. 10.3.1
and 10.3.2. It requires an FFAG ring lattice, radial or spiral, indifferently.
In the former case:
– a radial FFAG lattice is devised in Exercise 10.1, solutions are found 10.1;
– synchrotron acceleration in that lattice is devised in Exercise 10.5, solutions are
found 10.5.
Regarding the second type of lattice, spiral sector, similar synchronous accel-
eration simulations are performed in Exercise 10.13 (solutions 10.12). They resort
to the analytical field modeling FFAG-SPI, and to CAVITE[IOPT .= 6] as well for
synchronous acceleration.

References

1. V. Veksler, A new method of acceleration of relativistic particles. J. Phys. USSR 9, 153–158

(1945)
2. E.M. McMillan, The synchrotron. Phys. Rev. 68, 143–144 (1945)
3. L. Jones, F. Mills, A. Sessler et al., Innovation Was Not Enough (World Scientific, 2010)
4. J.R. Richardson, et al., Frequency modulated cyclotron. Phys. Rev. 69, 669 (1946);
J.R. Richardson, et al., Development of the frequency modulated cyclotron. Phys.
Rev. 73, 424 (1948). https://journals.aps.org/pr/abstract/10.1103/PhysRev.73.424 Univer-
sity of California, Berkeley, California, Lawrence Radiation Laboratory: The 184-inch
synchrocyclotron. https://www.gutenberg.org/files/33397/33397-h/33397-h.htm#THE_184-
INCH_SYNCHROCYCLOTRON
5. F.K. Goward, D.E. Barnes, Experimental 8 MeV synchrotron for electron acceleration. Nature
158, 413 (1946)
6. O.A. Shcherbakov, et al., Spallation neutron source at the 1 GeV synchrocyclotron of PNPI,
in Proceedings of the RuPAC-2016, Peterhof, St. Petersburg, 21–25 November, 2016. https://
accelconf.web.cern.ch/rupac2016/talks/wezmh01_talk.pdf Fig. 7.1: Copyrights under license
CC-BY-3.0, https://creativecommons.org/licenses/by/3.0; no change to the material
7. S. Henrotin, et al., Commissioning and testing of the first IBA S2C2, in TUP07 Proceed-
ings of Cyclotrons2016, Zurich, Switzerland. https://accelconf.web.cern.ch/cyclotrons2016/
papers/tup07.pdf Fig. 7.2: Copyrights under license CC-BY-3.0, https://creativecommons.org/
licenses/by/3.0; no change to the material
8. A. Laisné, et al., The Orsay 200 MeV Synchrocyclotron. https://accelconf.web.cern.ch/c78/
papers/a-09.pdf Fig. 7.3: Copyrights under license CC-BY-3.0, https://creativecommons.org/
licenses/by/3.0; no change to the material
9. S. Meyroneinc, et al., Beam quality for protontherapy at C.P.O, in Proceedings of the 15th
International Conference on Cyclotrons and their Applications, Caen, France (1998). https://
accelconf.web.cern.ch/c98/papers/a-04.pdf
10. P. Mandrillon, et al., Commissioning and implementation of the MEDICYC CYCLOTRON
PROGRAMME, in Proceedings of the Twelfth International Conference on Cyclotrons and
their Applications, Berlin, Germany. https://accelconf.web.cern.ch/c89/papers/e-05.pdf
236 7 Synchrocyclotron

11. F.T. Cole, O Camelot, a Memoir of the MURA Years. Cyclotron Conference, East Lansing,
USA, May 13–17 (2001). https://accelconf.web.cern.ch/accelconf/c01/cyc2001/extra/Cole.
pdf
12. M. Tanigaki, et al., Construction of FFAG accelerators in KURRI for ADS study, Proceedings
of the EPAC 2004 Accelerator Conference, pp. 2676–2678 (2004). http://accelconf.web.cern.
ch/accelconf/e04/PAPERS/THPLT078.PDF
13. L.M. Onischenko, JINR Phasotron, in Proceedings of the Pac 1987 Conference. https://
accelconf.web.cern.ch/p87/PDF/PAC1987_0878.pdf
14. F. Méot, Zgoubi Users’ Guide. https://www.osti.gov/biblio/1062013-zgoubi-users-guide.
Sourceforge latest version: https://sourceforge.net/p/zgoubi/code/HEAD/tree/trunk/guide/
Zgoubi.pdf

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0
International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing,
adaptation, distribution and reproduction in any medium or format, as long as you give appropriate
credit to the original author(s) and the source, provide a link to the Creative Commons license and
indicate if changes were made.
The images or other third party material in this chapter are included in the chapter’s Creative
Commons license, unless indicated otherwise in a credit line to the material. If material is not
included in the chapter’s Creative Commons license and your intended use is not permitted by
statutory regulation or exceeds the permitted use, you will need to obtain permission directly from
the copyright holder.
Chapter 8
Weak Focusing Synchrotron

Abstract This chapter introduces the weak focusing synchrotron, and the theoreti-
cal material needed for the simulation exercises. It begins with a brief reminder of the
historical context, and continues with the beam optics and acceleration techniques
that the weak focusing synchrotron principle and methods lean on, relying on basic
charged particle optics and acceleration concepts introduced in the previous chapters.
It further addresses the following aspects:

– fixed closed orbit,

– periodic structure,
– periodic motion stability,
– optical functions,
– synchrotron motion,
– depolarizing resonances.

The simulation of a weak focusing synchrotron lattice only requires two optical
elements: DIPOLE or BEND to simulate combined function dipoles, and DRIFT
to simulate straight sections. A third element, CAVITE, is required for accelera-
tion. Computation of synchrotron radiation (SR) Poynting and spectral brightness
uses zpop. Particle monitoring requires keywords introduced in the previous chap-
ters, including FAISCEAU, FAISTORE, PICKUPS, and some others. Spin motion
computation and monitoring resort to SPNTRK, SPNPRT and FAISTORE. Optics
matching and optimization use FIT[2]. INCLUDE is used, mostly here in order to
shorten the input data files. SYSTEM is used to, mostly, resort to gnuplot so as to
end simulations with some specific graphs (orbits, fields, or else) obtained by read-
ing data from output files such as zgoubi.fai (resulting from the use of FAISTORE),
zgoubi.plt (resulting from IL.= 2), or other zgoubi.*.out files resulting from a PRINT
command.

© The Author(s) 2024 237

F. Méot, Understanding the Physics of Particle Accelerators, Particle Acceleration
and Detection, https://doi.org/10.1007/978-3-031-59979-8_8
238 8 Weak Focusing Synchrotron

Notations Used in the Text

. B; .B; . Bx,y,s field; field vector; its components in the moving frame
. Bρ = p/q; Bρ0 particle rigidity; reference rigidity
[ straight
C; .C0
. orbit length,.C = 2π R + ; reference,.C0 = C( p = p0 )
sections
.E; . E σ , . E π SR electric field impulse; its parallel and normal components
. E; E s particle energy, . E = γ m 0 c2 ; synchronous energy
EFB Effective Field Boundary
. f rev , . f rf = h f rev revolution and RF voltage frequencies
.G gyromagnetic anomaly, .G = 1.792847 for proton
.h RF harmonic number, .h = f rf / f rev
.m; .m 0 ; . M particle mass; rest mass; mass in units of MeV/c.2
ρ ∂B
.n = − focusing index
B ∂x
.n0 stable spin precession direction
.p; . p; . p0 momentum vector; its modulus; reference
.P = E × B SR Poynting vector
. Pi , P f beam polarization, initial, final
.q particle charge
.R average orbit radius, . R = C/2π
.s path variable
.v particle velocity
. V (t); V̂ oscillating voltage; its peak value
x ,, y, horizontal and vertical coordinates in the moving frame
.α momentum compaction; or trajectory deviation; or depolarizing
resonance crossing speed
.β = v/c; .β0 ; .βs normalized particle velocity; reference; synchronous
β
. u betatron functions (.u : x, y)
.γ = E/m 0 c
2
Lorentz relativistic factor
.δp, ./\p momentum offset
.Ec critical energy of SR, .Ec = hωc = hc/λc
.ε wedge angle
.εu Courant-Snyder invariant; or beam emittance (.u : x, y, l)
.E R strength of a depolarizing resonance {
.μu betatron phase advance per period, .μu = period βuds(s) (.u : x, y)
.νu wave numbers, horizontal, vertical, synchrotron (.u : x, y, l)
.ρ; ρ0 curvature radius; reference
.σ beam matrix
.φ; φs particle phase at voltage gap; synchronous phase
.ϕ spin angle to the vertical axis
.ω angular frequency
.ωc critical angular frequency of SR, .ωc = 3γ 3 c/2ρ
8.1 Introduction 239

8.1 Introduction

The synchrotron is an outcome of the mid-1940s phase focusing resonant acceleration

concept [1, 2]. Phase focusing, or synchronous, acceleration with slow variation
of the magnetic field to maintain the beam on a constant orbit and constant RF
phase was demonstrated with the acceleration of electrons from 4 to 8 MeV, in an
existing betatron, using fixed RF, in 1946 [3]. This proof-of-principle was closely
followed by the construction and operation, at GEC, of a 70 MeV synchrotron (weak
focusing... no other choice at the time). The latter happened to be the opportunity
for the first observation of visible SR, a serendipity resulting from the fact that the
vacuum chamber was made of glass [4]. Observations included color of the radiation
changing from blue to yellow when energy was decreased to 40 MeV1 [5]—more in
the Poynting simulation Exercise 8.3. Measurements of properties of the radiation
were undertaken at the time, whereas SR acquired a status of a beam monitoring
tool, that was the beginning of a long story, still underway...
Transverse beam confinement in the weak focusing synchrotron version of the
synchrotron, over the thousands of turns needed for acceleration to top energy, was
based on the technique known at the time, inherited from cyclotron and betatron:
weak focusing,
Phase focusing states that stability of longitudinal motion (longitudinal focusing),
is obtained if the particles in a bunch arrive at the accelerating gap in the vicinity
of a proper phase of the oscillating voltage, the synchronous phase, such that the
bunch stays together during acceleration. Synchrotrons operate in general in a non-
isochronous regime: the revolution period changes with energy. As a consequence
the RF, . f rf = h f rev , has to change continuously from injection to top energy in order
to maintain an accelerated bunch on the synchronous phase. The reference orbit in
a synchrotron is maintained at constant radius by ramping the guiding field in the
main dipoles in synchronism with the acceleration, as in the betatron [6].
The synchrotron concept increased the energy reach of particle accelerators at the
time. It led to the construction of a series of proton rings with increasing energy [8]:
1 GeV at Birmingham (1953), 3.3 GeV at the Cosmotron (Brookhaven National Lab-
oratory, 1953–1969), 6.2 GeV at the Bevatron (Berkeley, 1954–1993), 10 GeV at the
Synchro-Phasotron (JINR, Dubna, 1957–2003), and a few others in the late 1950s.
Weak focusing magnets are quite bulky, creating a practical limit to further increase
in energy.2 This issue was overcome with the strong focusing method, devised in the
early 1950s (Chap. 9). The general layout of these first weak focusing synchrotrons
included straight sections (often 4, Figs. 8.1 and 8.2), to allow for the insertion of
injection and extraction systems, accelerating cavities, orbit correction and beam
monitoring equipment (Fig. 8.3).

1 At 70 MeV with a bending radius of say 0.5 m, the critical wavelength.λc = 4πρ/3γ 3 (Sect. 5.2.3)
falls in the visible range.
2 The story has it that it was possible to ride a bicycle in the vacuum chamber of Dubna’s Synchro-

Phasotron.
240 8 Weak Focusing Synchrotron

Fig. 8.1 SATURNE 1 at

Saclay [7], a 3 GeV,
4-period, 68.9 m
circumference, weak
focusing synchrotron,
constructed in 1956–1958.
The injection line is seen in
the foreground. Injection is
from a 3.6 MeV Van de
Graaff (not visible)

Fig. 8.2 A slice of the

SATURNE 1 dipole [7]. The
slight gap tapering,
increasing outward,
determines the weak index
condition .0 < n < 1

The weak focusing synchrotron was used in fixed-target nuclear and particle
physics, material science, medicine, industry, etc. Remarkably, it was a landmark (if
not the starting point3 ) of the history of collider rings, the AdA Anello di Accumu-
lazione, which demonstrated long term beam storage (and the Touschek effect), and

3 The third electron model built by the MURA group, a 50 MeV fixed field alternating gradient
(FFAG) ring, started in 1961, was operated in collider mode with two counter-rotating electron
beams [12, 13].
8.1 Introduction 241

Fig. 8.3 Loma Linda University medical synchrotron [9], during commissioning in 1989 at the
Fermilab National Accelerator Laboratory where it was designed

produced the first e+e- collisions in the early 1960s, was a weak focusing synchrotron,
a 250 MeV ring based on a .n = 0.55 gradient dipole [14].
Polarized beams
Synchrotrons allowed the acceleration of polarized beams to high energy.4 The
possibility was considered from the early times at Argonne ZGS (Zero-Gradient
Synchrotron), a 12 GeV weak focusing synchrotron operated over 1964–1979 [16]
(Fig. 8.4). ZGS accelerated polarized proton beams to 17.5 GeV/c with appreciable
polarization [17]. Polarization preservation techniques included harmonic orbit cor-
rection and fast betatron tune jumps at the strongest depolarizing resonances [18]
(cf. Sect. 8.2.4, Fig. 8.19). Experiments were performed to assess the possibility of
polarization transmission in strong focusing synchrotrons, and potential polarization
lifetime in colliders [19]. Acceleration of polarized deuteron was achieved in the late
1970s [20].
The weak focusing synchrotron is still topical today, due for a large part to its
relative simplicity, with low energy beam application where relatively low current is
not a concern, such as in the hadrontherapy (Fig. 8.3) [10, 11]. It only requires a single
type of a simple weak gradient dipole, a single power supply, a single accelerating
gap. It has an advantage of beam manipulation flexibility, when needed, compared
to (synchro-)cyclotrons.

4Polarized proton and deuteron beams had been accelerated in electrostatic columns (Sect. 2.1),
and soon after in cyclotrons, when polarized beam sources were made available.
242 8 Weak Focusing Synchrotron

Fig. 8.4 The ZGS at Argonne during construction [15]. A 12 GeV, 8-dipole, 4-period, 172 m
circumference, wedge focusing synchrotron. The two persons inside and outside the ring, in the
background, give an idea of the size of the magnets

8.2 Basic Concepts and Formulæ

The synchrotron is based on two key principles. First, a slowly varying magnetic
field maintains a constant orbit during acceleration,

. B(t) ρ = p(t)/q, ρ = constant, (8.1)

with . p(t) the particle momentum and .ρ the bending radius in the dipoles. Second,
longitudinal phase stability enables synchronous acceleration. In a regime where
velocity change with energy cannot be ignored (non-ultrarelativistic particles), the
latter requires a modulation of the accelerating voltage frequency to satisfy

f (t) = h f rev (t)

. rf with h an integer (8.2)

Synchronism between accelerating voltage oscillations and particle revolution keeps

the bunch on a synchronous phase. Synchronous acceleration is technologically sim-
pler in the case of electrons above a few MeV, because frequency modulation is
unnecessary. For instance, from .v/c = 0.9987 at 10 MeV to .v/c → 1 the relative
change in revolution frequency amounts to .δ f rev / f rev = δβ/β < 0.0013.
Varying field and RF on the one hand, fixed orbit in addition, are major evolu-
tions compared to cyclotron, where instead, field and RF are fixed, and the accelerated
orbit spirals out. A fixed orbit reduces the radial extent of individual guiding magnets,
8.2 Basic Concepts and Formulæ 243

allowing a structure comprised of a circular string of dipoles. A synchrocyclotron

instead uses a single, massive dipole (the volume of iron increases more than quadrat-
ically with bunch
{ rigidity) with a wide {radial extent allowing for a span of the field
integral over . Binjection dl = 2π qpmin − Bextraction dl = 2π qpmax .
Either a weak index (.−1 < k < 0, Sect. 3.2.2) and/or wedge focusing (cf.
Sect. 14.4.1) are used in weak focusing synchrotrons. Transverse stability was based
solely on the latter at Argonne ZGS. Weak focusing in the ZGS resulted in weak
depolarizing resonances, an advantage in that matter [19].
The synchrotron is a pulsed accelerator due to the necessary ramping of the field
in order to maintain a constant orbit. The acceleration is cycled, from injection to top
energy, repeatedly. The cycling repetition rate depends on the type of power supply.
If the ramping uses a constant electromotive force, then
[ ( ) ( )2 ]
− τt t t t
. B(t) ∝ (1 − e )=1− 1− + − ... ≈ (8.3)
τ τ τ

essentially linear. . Ḃ = d B/dt does not exceed a few Tesla/second, the repetition rate
of the acceleration cycle is of the order of a Hertz. If instead the magnet winding
is part of a resonant circuit then the field oscillates from an injection threshold to a
maximum value, . B(t) : B0 → B0 + B̂, as in the betatron. In this case the repetition
rate can be up to a few tens of Hertz. In both cases anyway B imposes its law and
the other quantities, RF frequency in particular, follow.
For comparison: in a synchrocyclotron the field is constant, thus acceleration
can be cycled as fast as the swing of the voltage frequency allows (hundreds of
Hz are common practice). A conservative 10 kV per turn requires of the order of
10,000 turns for a proton to reach 100 MeV, with velocity .0.046 < v/c < 0.43 from
1 to 100 MeV. Take .v ≈ c for simplicity, and a circumference of a few meters, the
acceleration thus takes .≈ 104 × C/c ≈ ms, potentially allowing a repetition rate in
the kHz range, more than an order of magnitude beyond the reach of a rapid-cycling
pulsed synchrotron.

8.2.1 Periodic Stability

This section introduces various ingredients concerning transverse focusing and the
conditions for periodic stability. It builds on material introduced in Chap. 3, Classical
Cyclotron.

Closed Orbit

The closed orbit is fixed, as in the betatron, and maintained during acceleration by
ensuring that the relationship of Eq. 8.1 is satisfied. In a perfect ring, the closed orbit
244 8 Weak Focusing Synchrotron

Fig. 8.5 A 4-fold symmetric

structure with four drift
spaces of length .2l. Orbit
length on reference
momentum . p0 is
.C = 2πρ0 + 8l. (O; s, x, y) is
the moving frame, along the
reference orbit. The orbit for
momentum . p = p0 + /\p
(./\p < 0, here) is at constant
distance ./\x = Dx /\pp0 from
the reference orbit

is along an arc in the bending magnets and straight along the drifts, Fig. 8.5. Particle
motion is defined in the Serret-Frénet frame (O; s, x, y), Fig. 3.8.

Transverse Focusing

Radial motion stability around a reference closed orbit in an axially symmetric dipole
field requires a field index (Sect. 3.2.2),
|
ρ0 ∂ B y ||
n=− (8.4)
B0 ∂ x |x=0, y=0
.

This quantity, evaluated on the reference arc in the dipoles, satisfies the weak focusing
condition (Eq. 3.12 with .n = −k)

. 0<n<1 (8.5)

This condition can be obtained with a tapered gap (as in SATURNE 1 dipole, Fig. 8.2)
resulting in both radial and axial focusing (Figs. 8.6 and 8.7). Note the sign convention
here, opposite to that used for the cyclotron (Eq. 3.11). This condition holds regardless
of the presence or not of drifts. Adding drifts brings to defining two radii, namely,
8.2 Basic Concepts and Formulæ 245

Fig. 8.6 Geometrical focusing: in a sector dipole with focusing index .n = 0, parallel incoming
rays of equal momenta experience the same curvature radius .ρ, so their trajectories converge as
outer trajectories have a longer path in the field. An index value n .= 1 cancels that effect: parallel
incoming rays exit parallel

Fig. 8.7 Axial motion

stability requires proper
shaping of field lines: . B y has
to decrease with radius. The
Laplace force pulls a positive
charge (located at . I ) with
velocity pointing out of the
page, toward the median
plane. Increasing the field
gradient (.n closer to 1, gap
opening up faster) increases
the focusing

(i) the magnet curvature radius .ρ0 ,

(ii) an average radius. R = C/2π = ρ0 + Nl/π (with.C the length of the reference
closed orbit, N the number of drifts and .2l their length) (Fig. 8.5) which can also be
written
Nl
. R = ρ0 (1 + k), k= (8.6)
πρ0

Adding drift spaces decreases the average focusing around the ring.
Geometrical focusing
The limit .n → 1 of the transverse motion stability domain corresponds to a cancel-
lation of the geometrical focusing (Fig. 8.6): in a constant field dipole (radial field
index n .= 0) the longer (respectively shorter) path in the magnetic field for paral-
lel trajectories entering the magnet at greater (respectively smaller) radius result in
convergence. This effect is cancelled (i.e., the bend angle is the same whatever the
entrance radius) if the curvature center is made independent of the entrance radius:
246 8 Weak Focusing Synchrotron

Fig. 8.8 Left: a focusing wedge (.ε < 0); opening the sector increases horizontal focusing and
decreases vertical focusing. Right: a defocusing wedge (.ε > 0), closing the sector, has the reverse
effect. This is the origin of the focusing in the ZGS zero-gradient dipoles

. O O , = 0, . O ,, O = 0. This occurs if trajectories at an outer (inner) radius experi-

ence a smaller (greater) field such as to satisfy . B L = Bρ = C st . Differentiating
. Bρ = C
st
gives . /\B
B
+ /\ρ
ρ
= 0, with ./\ρ = /\x, so yielding .n = − ρB00 /\B
/\x
= 1. The
ρ02
focal distance associated with the curvature is (Eq. 3.13 with . R = ρ0 ) . f = L
.
Wedge Focusing
Entrance and exit wedge angles may be used to ensure transverse focusing, Fig. 8.8:
opening the magnetic sector increases the horizontal focusing (and decreases the
vertical focusing); closing the magnetic sector has the reverse effect (cf. Sect. 14.4.1).
At the wedge the trajectory undergoes a deviation proportional to the distance to
the optical axis, amounting to

tan ε tan(ε − ψ)
/\x , =
. /\x, /\y , = − /\y (8.7)
ρ0 ρ0

The angle .ψ is a correction for the fringe field extent (Eq. 14.20); the effect is of the
first order on the vertical focusing, and second order horizontally.
Profiling the magnet gap in order to adjust the focal distance complicates the
magnet; a parallel gap, .n = 0, makes it simpler, for that reason edge focusing may
be preferred. The method benefited the acceleration of polarized beams in the ZGS, as
radial field components (which are responsible for depolarization), met at the EFBs
of the eight main dipoles, were therefore weak [17]. Preserving beam polarization at
high energy required tight control of the tunes, achieved at the 0.01 level by means
of pole face winding added at the ends of the dipoles [21, 22].
Drawbacks of the weak focusing method include interdependence of radial and
axial focusing, see Working point Section, below.
8.2 Basic Concepts and Formulæ 247

Betatron Motion

The first order differential equations of motion in the moving frame (Fig. 8.5) derive
from the Lorentz equation
⎧ ds ⎫ ⎧( ) ⎫
⎨ s ⎬ ⎪
⎨ dx
B y − dy
B x s⎪
⎬
dmv d dt dt dt
. = qv × B ⇒ m dx
x = q − dt B y x
ds (8.8)
dt dt ⎩ dy
dt ⎭ ⎪
⎩ ⎪
⎭
dt
y ds
Bx y dt

Motion in a weak index dipole field is solved in Sect. 3.2.2, Classical Cyclotron
∂B
chapter: in Eq. 3.7 substitute .ρ to . R, .n = − ρB00 ∂ xy to .−k (Eq. 3.11), and evaluate
on the reference orbit. Taylor expansions of the transverse field components in the
moving frame lead to

. B y (ρ)|y=0 = B0 (1 − n ρx0 ) + O(x 2 )

Bx (0 + y) = −n ρB00 y + O(y 3 ) (8.9)

Assume transverse stability: .0 < n < 1. In the approximation .ds ≈ vdt (Eq. 3.14)
Eqs. 8.8, and 8.9 lead to the differential equations of motion

d2x 1−n d2 y n
. + x = 0, + 2y =0 (8.10)
ds 2 ρ02 ds 2 ρ0

In an periodic structure comprised of gradient dipoles, wedges and drift spaces, the
differential equation of motion takes the general form of Hill’s equation, namely
(with .u standing for .x or . y),
⎧ {
⎧ 2 ⎪
⎪ K x = 1−n
⎪
⎪ in dipoles : ρ02
⎨ d u + K (s)u = 0 ⎨ K y = ρ2
n
u
ds 2 with 0
= s0 : K x = ± ρtan0 ε δ(s − s0 )
.
⎩ ⎪
⎪ at a wedge at s
K u (s + S) = K u (s) ⎪
⎪ y
⎩ in drift spaces : 1 = 0, K = K = 0
ρ0 x y
(8.11)
Here . K u (s) is periodic, . S = 2π R/N (. S = C/4 for instance in a 4-period ring,
Figs. 8.1 and 8.5).
The solution of Eqs. 8.11 is not as straightforward as in the cyclotron where a
constant . K u around the ring (Eq. 3.15) results in a sinusoidal motion (Eq. 3.17). A
sinusoidal motion, with adding drifts, however remains a reasonable approximation,
see below, Weak focusing approximation.
Floquet established [23] that the two independent solutions of Hill’s second order
differential equation with periodic coefficient have the form [24]
248 8 Weak Focusing Synchrotron
⎧ { s ds
⎪
⎪ i 0 {
⎨ √ βu (s)
u 1 (s) = βu (s) e u 2 (s) = u ∗1 (s)
and (8.12)
du 2 (s)/ds = du ∗1 (s)/ds
.
⎪
⎪ i − α (s)
⎩ du 1 (s)/ds = u
u 1 (s)
βu (s)

where .βu (s) and .αu (s) = −βu, (s)/2 are periodic functions, from what it results that

{s ds
±i
.u 1 (s + S) = u 1 (s) e
s0
βu (s) (8.13)
2 2

{ s ds
where . s0 is the betatron phase advance at .s, from the origin .s0 . A real
βu (s)
solution of Hill’s equation is the linear combination . A u 1 (s) + A∗ u ∗2 (s). With
1√
.A =
2
εu /π eiφ following conventional notations, .φ the phase of the motion at the
origin .s = s0 , the general solution of Eq. 8.11 is
⎧ ( )
⎪ √ { s ds
⎪
⎨ u(s) = βu (s)εu /π cos s0 +φ
/ ( βu ) ( )
.
{ s ds { s ds (8.14)
⎪
⎪ , εu /π
⎩ u (s) = − sin s0 + φ + αu (s) cos s0 +φ
βu (s) βu βu

The Courant-Snyder invariant of the motion is

εu 1 [ 2 ( )2 ]
. = u + αu (s)u + βu (s)u , (8.15)
π βu (s)

At a given azimuth .s of the periodic structure the observed turn-by-turn motion lies
on that ellipse (Fig. 8.9). The form and orientation of the ellipse feature a weak
dependence on the observation azimuth .s, via the respective local values of .αu (s)
(small at all .s) and .βu (s) (weakly modulated), and its area .εu is an invariant. Equa-
tion 8.14 taken for .αu (s) = 0 (an observation azimuth .s where the ellipse is upright)
shows
( that motion along ) the ellipse is clockwise. Note that in the coordinate system
,
.(u, αu (s)u + βu (s)u ) the particle moves on a circle of radius .εu /π .
The phase advance over a turn (from one position to the next on the ellipse,
Fig. 8.9) in an N-periodic ring yields the wave number
{ s0 +N S {
1 ds N ds N μu
.νu = = = (8.16)
2π s0 βu (s) 2π period βu (s) 2π
8.2 Basic Concepts and Formulæ 249

Fig. 8.9 A thousand passes in a ZGS 43 m cell, observed at the center of the long drift where
.αx (s)= 0, materialize the upright horizontal Courant-Snyder invariant. The first five passes are
marked, motion goes clockwise with a cell phase advance of .0.21 × 2π . The aspect ratio of the
ellipse only weakly depends on .s, its area (.εx = 100π μ m here) is an invariant of the motion

Weak focusing approximation

In a cylindrically symmetric structure the sinusoidal motion is the exact solution of the
first order differential equations of motion (Eqs. 3.16 and 3.17, Classical Cyclotron
chapter), the coefficients . K x = (1 − n)/ρ02 and . K y = n/ρ02 are independent of .s.
Adding drift spaces results in Hill’s differential equation with periodic coefficient
. K (s + S) = K (s) (Eq. 8.11), with solution a pseudo harmonic motion (Eq. 8.14).
Due to the weak focusing the beam envelope is only weakly modulated (cf. below),
thus also is .βu (s). In practice the modulation of .βu (s) does not exceed a few percent,
justifying the introduction of the average value .β u to approximate the phase advance
by { s
ds s s
. ≈ = νu (8.17)
0 βu (s) β u R

The right equality is obtained by applying this approximation to the phase advance
per period, namely
{ s0 +S
ds S
.μu = ≈ (8.18)
s0 βu (s) βu

and introducing the wave number of the N-period optical structure (Eq. 8.16) so that

R
β =
. u (8.19)
νu
250 8 Weak Focusing Synchrotron

the wavelength of the betatron oscillation. With .k << 1 and using Eq. 8.23,

ρ0 (1 + k/2) ρ0 (1 + k/2)
. xβ = √ , βy = √ (8.20)
1−n n
{
Substituting .νu Rs to . ds
βu (s)
in Eq. 8.14 yields the approximate solution
⎧ √ ( s )
⎪
⎨ u(s) ≈ βu (s)εu /π cos νu + φ
/ ( s R ) ( s ) (8.21)
.
⎪ , εu /π
⎩ u (s) ≈ − sin νu + φ + αu (s) cos νu + φ
βu (s) R R

Beam envelopes
The beam envelope .û(s) (with .u standing for .x or . y) is determined by a particle on
the maximum invariant .εu /π . It is given at all .s by
/
εu
.û(s) = ± βu (s) (8.22)
π

As .βu (s) is . S-periodic, so also is the envelope, .û(s + S) = û(s). In a cell with
symmetries, the beam envelopes feature the same symmetries, as shown in Fig. 8.10

Fig. 8.10 Multiturn particle excursion (absolute values, .|x(s)| and .|y(s)|) along the ZGS 2-dipole
43 m cell. The motion extrema (.[βu (s)εu /π ]1/2 , Eq. 8.22) tangent the envelops, respectively hor-
izontal (red, across the dipoles), and vertical (blue). Envelops are only weakly modulated. They
feature the symmetry of the cell
8.2 Basic Concepts and Formulæ 251

for the ZGS: a symmetry with respect to the center of the cell. Envelope extrema
are at azimuth .s of .βu (s) extrema, i.e. where .d û(s)/ds ∝ βu, (s) = 0 or .αu = 0 as
,
.βu = −2αu .

Working point
The “working point” of the synchrotron is the wave number pair .(νx , ν y ) at which
the accelerator is operated, it fully characterizes the focusing. √ In a structure with
√
cylindrical symmetry (such as the classical cyclotron) .νx = 1 − n and .ν y = n
(Eq. 3.18) so that .νx2 + ν y2 = 1: when the radial field index .n is changed the work-
ing point stays on a circle of radius 1 in the stability diagram (or “tune diagram”,
√
Fig. 8.11). If drift spaces are added, from Eqs. 8.19 and 8.20, with .1 + k2 ≈ R/ρ0
(Eq. 8.6), it comes
/ /
R R R
ν ≈
. x (1 − n) , ν y ≈ n , νx2 + ν y2 ≈ (8.23)
ρ0 ρ0 ρ0
√
Thus the working point is located on a circle of radius . R/ρ0 > 1 (Fig. 8.11), tunes
can not exceed the limits /
.0 < νx, y < R/ρ0

Horizontal and vertical focusing are not independent (Eq. 8.11): if .νx increases
.
then .ν y decreases and vice versa. This is a lack of flexibility which strong focusing
overcomes by providing two knobs allowing separate adjustment.

Fig. 8.11 Location of the

working point in the tune
diagram. (A) field with
revolution symmetry:
.(νx , ν y ) is on a circle of
radius 1; (B) sector field with
index .0 < n < 1 and drift
spaces: .(ν√ x , ν y ) is on a circle
of radius . R/ρ0 ; (C) strong
focusing, AG index .|n| >> 1
or separated function: .νx and
.ν y are large, set
independently
252 8 Weak Focusing Synchrotron

Fig. 8.12 In a sector dipole

with radial index .n /= 0,
closed orbits follow arcs of
constant field. A closed orbit
at . p0 + /\p follows an arc of
radius .ρ0 + /\ρ,
./\ρ = /\p/(1 − n)q B0

Off-momentum orbits; periodic dispersion

In the linear approximation in ./\p/ p0 , a momentum offset ./\p = p − p0 changes
mv to .mv(1 + /\p/ p0 ) in Eq. 8.8. This changes the horizontal equation of motion
.

(Eq. 8.10) to
( )
d2x 1 /\p d2x 1 /\p
. + Kx x = , or + Kx x − =0 (8.24)
ds 2 ρ0 p0 ds 2 ρ0 K x p0

A change of variable .x − K x1ρ0 /\p

p0
→ x (with .1/ ρ0 K x = ρ0 /(1 − n)) restores the
unperturbed equation of motion; thus orbits of different momenta . p = p0 + /\p are
separated by
ρ0 /\p
./\x = (8.25)
1 − n p0

from the reference orbit (Fig. 8.12). Introducing the geometrical radius . R = (1 +
k)ρ0 (Eq. 8.6) to account for the added drifts, this yields the dispersion function

/\x /\R R ρ0
. Dx = ≡ = = , constant, positive
/\p/ p0 /\p/ p0 (1 − n)(1 + k) 1−n
(8.26)
. D x is the chromatic dispersion of the orbits, an s-independent quantity: in a structure

with axial symmetry, comprising drift sections (Fig. 8.5) or not (classical and AVF
cyclotrons for instance), the ratio ./\x / /\p/ p0 is independent of the azimuth .s, the
distance of a chromatic orbit to the reference orbit is constant around the ring.
Given that .n < 1,
– higher momentum orbits, . p > p0 , have a greater radius,
– lower momentum orbits, . p < p0 , have a smaller radius.
The horizontal motion of an off-momentum particle is a superposition of the
betatron motion (solution of Hill’s Eq. 8.21 taken for .u = x) and of a particular
solution of the inhomogeneous equation (Eq. 8.24), namely
8.2 Basic Concepts and Formulæ 253

/ ( s ) ρ0 /\p
. x(s) = βu (s)εu /π cos νu + φ + (8.27)
R 1 − n p0

The vertical motion is unchanged.

Chromatic orbit length
In an axially symmetric structure the difference in closed orbit length ./\C = 2π /\R
resulting from the difference in momentum comes from the dipoles, as all orbits
are parallel in the drifts (Fig. 8.5). Hence, from Eq. 8.26, the relative closed orbit
lengthening factor, or momentum compaction, is

/\C / /\p /\R / /\p 1 1

α=
. ≡ = ≈ 2 (8.28)
C p0 R p0 (1 − n)(1 + k) νx

with .k = Nl/πρ0 (Eq. 8.6). Note that the relationship .α ≈ 1/νx2 between momen-
tum compaction and horizontal wave number established for a revolution symmetry
structure (Eq. 3.22) still holds when adding drifts.

8.2.2 Acceleration

The field . B in a synchrotron is varied during acceleration (a function performed by

the magnet power supply) concurrently with the variation of the bunch momentum
. p (a function performed by the accelerating cavity) in such a way that the beam
stays on the design orbit. Given the energies involved, the magnet supply imposes its
law . B(t) (Fig. 8.13), and the cavity follows the best it can. The accelerating voltage
. V̂ (t) sin ωrf t is maintained in synchronism with the revolution motion by ensuring,
as well as possible,
c B(t)
ωrf = hωrev = h /
R ( m c )2
.

0
qρ
+ B 2 (t)

Typically, for a.C = 2π R ≈ 70 m circumference ring,5 accelerating from.β = v/c ≈

.
0.09 at injection (3.6 MeV protons) to .β ≈ 1 at top energy (3 GeV), the revolution
period .Trev = C/βc and frequency .ωrev /2π = 1/Trev span
{
Trev : 2.6 µs → 0.24 µs
.
f rev : 380 kHz → 4.2 MHz

5 Case of the SATURNE 1 weak focusing synchrotron (Fig. 8.1), cf. Exercise 8.1, Table 8.1
254 8 Weak Focusing Synchrotron

Energy gain
The variation of the particle energy over one turn amounts to the work of the force
F = dp/dt = qρd B/dt on the charge at the cavity, namely
.

/\W = F · 2π R = 2π Rqρ Ḃ
. (8.29)

In a slow-cycling synchrotron . Ḃ is usually constant over most of the acceleration

cycle (Eq. 8.3), and so is ./\W . At SATURNE 1, for instance

/\W
. = 2π Rρ Ḃ = 68.9[m] × 8.42[m] × 1.8[T/s] = 1044 volts/turn
q

The field ramp lasts

/\t = (Bmax − Bmin )/ Ḃ ≈ Bmax / Ḃ = 0.8 s

.

The number of turns to the top energy (.Wmax ≈ 3 GeV) is

Wmax 3 109 eV
. N= = ≈ 3 × 106 turns
/\W 1044 eV/turn

The dependence of particle mass on field writes

.

/( )2
qρ m0c
.m(t) = γ (t)m 0 = + B 2 (t)
c qρ

Fig. 8.13 Cycling . B(t) in a pulsed synchrotron. Ignoring saturation, . B(t) during the ramp is
proportional to the magnet power supply current . I (t). Beam injection occurs at low field, in the
region of A, while extraction occurs at top energy on the high field plateau. (AB): field ramp up
(acceleration); (BC): flat top; (CD): field ramp down; (DA’): thermal relaxation. (AA’): repetition
period; (1/AA’): repetition rate; slope : ramp velocity . Ḃ = d B/dt (T/s)
8.2 Basic Concepts and Formulæ 255

Fig. 8.14 Adiabatic damping of betatron oscillations from .u , = pu / ps to .u ,2 = pu /( ps + /\ps ) at

( ) cavity. In transverse phase space the particle motion invariant .εu decreases, as a
the accelerating
result of ./\ du
ds

Adiabatic damping of the betatron oscillations

Particle momentum increases at the accelerating gap, resulting in a decrease of the
amplitude of betatron oscillations (an increase if deceleration). The mechanism
is sketched in Fig. 8.14 (the solution of the equations of motion is addressed in
Sect. 10.2.3). The slope at the cavity is
| |
m du du || m du |
du pu dt | pu,2
.before the cavity: = ds
dt
= , after: = ds ||
=
ds m dt ps ds |2 m dt ps,2
2

with.u standing for.x or. y. As the kick in momentum is longitudinal,.dpu /dt = 0 thus
.

pu,2 = pu and the increase in momentum is purely longitudinal, . ps,2 = ps + /\ps .

.
Thus |
du || pu pu /\ps
= ≈ (1 − )
ds |2
.
ps + /\ps ps ps

and as a consequence the slope .du/ds varies across the cavity,

( ) |
du du || du du /\ps
/\ = − =− , proportional to the slope
ds |2 ds
.
ds ds ps

If ./\p/ p > 0 (acceleration) then the slope decreases. This variation has two conse-
quences on the betatron oscillation (Fig. 8.14):
– a change of the betatron phase,
– a modification of the betatron amplitude.
Coordinate transport
At the cavity {
u2 = u
u ,2 ≈ ppus (1 − = u , (1 −
. dp
p
) dp
p
)
256 8 Weak Focusing Synchrotron

In matrix form,
( ) ( ) [ ]
u2 u 1 0
= [C] with [C] = (8.30)
u ,2 u,
.
0 1 − dp
p

Since .det[C] = 1 − dpp

/= 1 the system is non-conservative and the area of the beam
ellipse in phase space is not conserved. Assume one cavity in the ring and note
.[T ] × [C] the one-turn coordinate transport matrix with origin at entrance of the
cavity. Its determinant is

dp
det[T ] × det[C] = det[C] = 1 −
.
p

The variation of the transverse ellipse area satisfies .εu = (1 −

.
dp
p
)ε0 or, with .dεu =
εu − ε0 , . dε
εu
u
= − dp
p
, The solution is

. p εu = constant, or βγ εu = constant (8.31)

Over . N turns the coordinate transport matrix is .[TN ] = ([T ][C]) N , thus the ellipse
area changes by a factor

dp N dp
det[C] N = (1 −
. ) ≈1− N
p p

Phase stability
The motion of a particle in the longitudinal phase space .(φ, δp/ p) is stabilized in
the vicinity of a synchronous phase, .φs , by the mechanism of phase stability, or
longitudinal focusing (Fig. 8.15). It requires
(i) the presence of an RF cavity with frequency locked on the revolution time,
(ii) bunch centroid to be positioned either on the rising slope of the oscillating
voltage (low energy regime), or on the falling slope (high energy regime).
The synchronous particle follows the reference closed orbit, its velocity satisfies
.v(t) =
q Bρ(t)
m
. At each turn it reaches the accelerating gap when the oscillating voltage
is at the synchronous phase .φs , and undergoes an energy gain

. /\W = q V̂ sin φs

The condition.| sin φs | < 1 imposes a lower limit to the cavity voltage for acceleration
to happen. According to Eq. 8.29,

. V̂ > 2π Rρ Ḃ
8.2 Basic Concepts and Formulæ 257

Fig. 8.15 A sketch of the mechanism of phase stability, .h = 3 in this example. Below transition
phase stability occurs for a synchronous phase taken at either one of A, A’, A” arrival times at the
gap. Beyond transition the stable phase is at either one of B, B’, B’ locations

Referring to Fig. 8.15, the synchronous phase can be placed on the left (A, A’, A” ...
series) or on the right (B, B’, B” ... series) of the oscillating voltage crest. One and
only one of these two possibilities, and which one depending upon the optical lattice
and on particle energy, ensures that particles in a bunch remain grouped in the vicinity
of the synchronous particle.
The transition is between these two time-of-flight regimes. Consider a particle
with higher energy compared to the synchronous particle:
– if the increase in path length around the ring is faster than the increase in veloc-
ity (case of classical cyclotron and synchrocyclotron; and of high energy electron
synchrotron, where velocity essentially does not change), a revolution around the
ring takes more time, the particle arrives at the accelerating gap late (.φ(t) > φs ); in
order for it to be pulled toward bunch center (i.e., take less time around the ring) it
has to lower its energy increase; this is the B series, above transition;
– if the velocity increase is faster than the path length increase (case in general of
synchrotrons at low energy), revolution around the ring takes less time, the particle
arrives at the accelerating gap early (.φ(t) < φs ); in order for it to be pulled toward
bunch center (i.e., take more time around the ring) it has to lower its energy increase;
this is the A series, below transition.
Transition energy
dTrev
The transition between the two time-of-flight regimes occurs when . = 0. With
Trev
. T = 2π/ω = C/v, this can be written

dωrev dTrev dv dC
. =− = −
ωrev Trev v C
258 8 Weak Focusing Synchrotron

dC dp
With . dv = 1 dp
γ2 p
and momentum compaction .α = / p , (Eq. 8.28), it becomes
v
C
( )
dωrev dTrev 1 dp dp
. =− = −α =η (8.32)
ωrev Trev γ 2 p p

which introduces the phase slip factor

kinematics
,,,,
1 1 1
.η = − ,,,,
α = 2− 2 (8.33)
γ 2 γ γtr
lattice

The “transition gamma”, .γtr , is a property of the lattice.

In a weak focusing lattice, after Eq. 8.28 and classical cyclotron’s Eq. 3.22,
√
γ = 1/ α ≈ νx
. tr (8.34)

Thus the phase stability regime is

below transition, i.e. φs < π/2, if γ < νx

.

above transition, i.e. φs > π/2, if γ > νx (8.35)

√
In a weak focusing synchrotron the horizontal tune .νx = (1 − n)R/ρ0 (Eq. 8.23)
may be . >
< 1, and subsequently .γtr > 1 is a possibility, .γtr may have to be crossed
during acceleration. There is no transition gamma if .νx < 1. At SATURNE 1 for
instance, with.νx ≈ 0.7 (Table 8.1) and.γtr < 1. So, ramping in energy did not require
crossing transition-gamma.6

8.2.3 Synchrotron Radiation Poynting

Visible SR was first observed in the GEC 70 MeV weak focusing synchrotron [4].
So, the bases of SR theory may opportunistically be recalled here [26, 27]. This
theoretical material serves the purpose of the exercises in addition. The topic is
further explored in Sect. 9.2.6, which addresses some aspects of the use of visible
SR for high energy electron or proton beam imaging.
In addressing low energy SR, the Poynting vector

P =E×B
.

6Transition-.γ crossing (Sect. 8.2.2) is a common longitudinal phase space beam manipulation
during acceleration in strong focusing synchrotrons. It requires an RF phase jump [25].
8.2 Basic Concepts and Formulæ 259

Fig. 8.16 The frame and vectors entering in the definition of the electric field radiated by the
accelerated particle (Eq. 8.36). Zgoubi notations are used here (Fig. 1.3): .(X, Y ): horizontal
plane; . Z : vertical axis; .R(t) is the particle position in the laboratory frame .(O, X, Y, Z ). Besides,
.u is the position of the observer; .r(t) = u − R(t) is the position of the particle with respect to the
observer; .n(t) = r(t)/|r(t)| is the (normalized) direction of observation; .β = (1/c)dR/dt is the
normalized velocity vector of the particle

is the relevant quantity [27, 28]. The electromagnetic field is given by the Liénard-
Wiechert equations in the long distance approximation

q n(t) × [(n(t) − β(t)) × dβ/dt] 1

.E(n, τ ) = , B= n×E (8.36)
4π ε0 c r (t) (1 − n(t) · β(t))3 c

where .n = r/r is the direction of observation (Fig. 8.16), .β = v/c, .β̇ = dβ/dt, .t
is the “retarded time”, at which the particle emitted the radiation, .τ is the observer
time, a little later. Namely, when at position .r(t) with respect to the observer, the
particle emits a signal which reaches the observer at time

τ = t + r (t)/c
. (8.37)

Electric impulse, From Raytracing [29, Sect. 3.2.1]

The vectors .n, .β, .β̇ are sub-products of the stepwise integration of particle motion.
They are used to compute .E(n, τ ) in zpop (its subroutine sref.f, actually).
The electric field impulse .E(n, τ ) is decomposed into two polarization compo-
nents . E σ (n, τ ) and . E π (n, τ ), respectively parallel and normal to the bend plane
(Fig. 8.17).
As an example, results for GEC 70 MeV synchrotron are given in Fig. 8.17:
the electric field impulses have been derived from the electron trajectory obtained
by stepwise raytracing, using Eqs. 8.36 and 8.37. The spectral brightness follows,
Eqs. 8.38 and 8.39.
260 8 Weak Focusing Synchrotron

Fig. 8.17 Left: typical shape of. E σ, π (τ ) impulses as observed in the direction.φ = 0, ψ ≈ 0.1/γ ,
in GEC 70 MeV synchrotron (by comparison, at.φ = ψ = 0,. E σ (τ )) is marginally different, whereas
. E π (τ ) ≡ 0). Right: spectral brightness, peaking near .hωc = 2γ c/2ρ ≈ 2.7 eV (.λc = 0.47 µm);
3

at such small .ψ ≈ 0.1/γ , the .π component of the radiation (blue curve) is quite weak compared to
the .σ component (red curve)

Spectral brightness [29, Sect. 3.2.2]

The respective Fourier transforms of the electric field impulse components . E σ (n, τ )
and . E π (n, τ ), namely
{ √
. Ẽ σ, π (φ, ψ, ω) = E σ, π (φ, ψ, t)e−iωτ dτ / 2π (8.38)

provide the spectral angular brightness (Fig. 8.17)

∂ 3 Pσ, π | |2
| |
. = 2E0 cr 2 | Ẽ σ, π (φ, ψ, ω)| (8.39)
∂φ∂ψ∂ω

with .φ and .ψ the angles of .n to respectively the .(X, Z ) and .(X, Y ) planes. Zpop
∂ 3 Pσ, π
computes . Ẽ σ, π (φ, ψ, ω) and . ∂φ∂ψ∂ω (its subroutine srdw.f, actually).
Electric impulse, Analytical [27] [29, Sect. 3.2]
The following theoretical reminders are resorted to in the exercises, for instance for
comparison to numerical outcomes from the computation of the electric impulse
.E(n, τ ) (Eq. 8.36) from numerical integration. Referring to Fig. 8.16, the observer
direction and velocity vectors write

n = (cos ψ cos φ, cos ψ sin φ, sin ψ), β = β(cos ω0 t, sin ω0 t, 0)

. (8.40)

The observer time .τ and, to order .1/γ 2 , its differential element are obtained from
the particle time .t (the numerical integration time) using

1 + γ 2ψ 2 ω02 3 dτ 1 + γ 2ψ 2 1
τ=
. t + t , = 1 − n(t) · β(t) _ + (ω0 t − φ)2
2γ 2 6 dt 2γ 2 2
(8.41)
8.2 Basic Concepts and Formulæ 261

with .ω0 ≈ c/ρ and .ρ the local curvature radius. The origin of observer time is at
φ = ψ = 0, i.e. in the direction tangent to particle trajectory. The radiated electric
.
impulses result, namely
( )
qω0 γ 4 (1 + γ 2 ψ 2 ) − γ 2 (ω0 t − φ)2 t
. E σ (t) = ( )3 rect
π E0 cr 1 + γ ψ + γ (ω0 t − φ)
2 2 2 2 2T
qω0 γ 4 −2γ ψγ (ω0 t − φ) t
E π (t) = ( )3 rect( ) (8.42)
π E0 cr 1 + γ ψ + γ (ω0 t − φ)
2 2 2 2 2T

where .rect(x) = 1 if − 21 < x < 21 , zero otherwise, defines the boundary of the
numerical integration, namely over a particle deviation angle .α = ±ω0 T . The
impulse components . E σ,π (t) in particle time have similar shapes to . E σ,π (τ ) at the
observer, Eq. 8.43 (Fig. 8.17), they differ in width as the latter is squeezed according
to .τ (t) contraction (Eq. 8.41)—a squeeze resulting from a double Doppler shift from
electron trajectory arc in the lab, to electron frame, and back to observed radiation
in the lab, ./\τ ≈ /\t/γ 2 . The cubic dependence .τ (t) (Eq. 8.41) has an analytical
solution; substitution of that solution .t (τ ) in Eq. 8.42 yields analytical expressions
for the field impulses in observer time,

qω0 γ 4 1 − 4 hsin2 [ 13 Ahsin u(φ, ψ, τ )] τ

. E σ (φ, ψ, τ ) = rect[ ] (8.43)
π E0 cr (1 + γ 2 ψ 2 )2 (1 + 4 hsin2 [ 13 Ahsin u(φ, ψ, τ )])3 2l(φ, ψ)
qω0 γ 4 4γ ψ hsin[ 13 Ahsin u(φ, ψ, τ )] τ
E π (φ, ψ, τ ) = rect[ ]
π E0 cr (1 + γ 2 ψ 2 )5/2 (1 + 4 hsin2 [ 13 Ahsin u(φ, ψ, τ )])3 2l(φ, ψ)

where .l(φ, ( ψ) is 2the )observation direction dependent signal duration and .u =

1 √ γφ γ φ2
2
3 + 1+γ 2 ψ 2 − 2 (1+γ ω2 ψc 2 )3/2 τ . The critical frequency .ωc partitions the
1+γ 2 ψ 2
{ω {∞
power spectrum in two equal parts, . 0 c ddωP = ωc ddωP . Equation 8.43 can be used
to check numerical integration outcomes. Refer to [26, 27, Sect. 4.4] for additional
details.
The typical width of the impulse in observer time is the familiar

1 ( )3/2 3γ 3 c
/\τc = ±
. 1 + γ 2ψ 2 with ωc = the critical frequency (8.44)
ωc 2ρ

Accounting for the.γ 2 double Doppler shift contraction,./\τc thus corresponds to a

trajectory arc length .lc ≈ c/\τc × γ 2 ≈ ρ/γ . From the point of view of the observer,
the SR power, integrated over the all spectrum, is mostly contained in an rms opening
angle [26, Sect. 5.5.2] (Fig. 8.18)

1
./\φc,rms = 0.83 (8.45)
γ
262 8 Weak Focusing Synchrotron

Fig. 8.18 Electron trajectory in the lab frame .R (right) and in .R, traveling parallel to, and at, its
velocity (left). The radiation, spanning a.± pi/2 angle in.R, , is confined in a forward cone of opening
.±/\φc ≈ ±1/γ in .R

For a given observation frequency.ω, the rms opening angle is a function of frequency
and satisfies [26, Sect. 5.5.1]

1 ( ωc )1/3
/\φc (ω) ≈ 0.72
. (8.46)
γ ω

Lower radiation energy has wider opening.

8.2.4 Depolarizing Resonances

The field index is zero in the ZGS, transverse focusing is ensured by wedge angles
at the ends of the eight dipoles, the only locations where non-zero horizontal field
components are present. The latter are weak and as a consequence so also are depo-
larizing resonances: “As we can see from the table, the transition probability [from
spin state .ψ1/2 to spin state .ψ−1/2 ] is reasonably small up to .γ = 7.1” [17], i.e.
proton .Gγ = 12.73, . p = 6.6 GeV/c. The table referred to stipulates a transition
probability . P 1 ,− 1 < 0.042, whereas resonances beyond that energy range feature
2 2
. P 1 ,− 1 > 0.36. Beam depolarization up to 6 GeV/c, under the effect of these reso-
2 2
nances, is illustrated in Fig. 8.19.
In a synchrotron using gradient dipoles, particles experience radial fields . Bx (y) =
−n ρB00 y as they undergo vertical betatron oscillations [17, 30, 31]. As .n is small these
radial field components are weak, and so is their effect on spin motion.
In a P-periodic ring, the vertical betatron motion excites “systematic intrinsic”
spin resonances, located at

. Gγ R = k P ± ν y , k∈N

If the P periodicity of the optics is lost (due to an optical defect), all resonances,
systematic and non-systematic, .Gγ R = integer ± ν y are excited. In the ZGS for
8.2 Basic Concepts and Formulæ 263

Fig. 8.19 Polarization loss at the ZGS [33] through the strong intrinsic resonances .Gγ R = 7.2
(. p = 3.65 GeV/c) and 8.8 (.4.51 GeV/c) (black circles). A vertical tune jump method preserves
polarization (empty circles)

instance, .ν y ≈ 0.8 (Table 8.2), the ring is . P = 4-periodic, thus .Gγ R = 4k ± 0.8.
Strongest intrinsic resonances are located at

. Gγ R = k m P ± ν y

with m the number of cells per superperiod [32, Sect. 3.II]. In the ZGS, with m .= 2
the strongest resonances occur at (Fig. 8.19)

. Gγ R = 2 × 4k ± 0.8 = 7.2 (3.65 GeV/c); 8.8 (4.51 GeV/c); 15.2 (7.9 GeV/c); ...

In the presence of vertical orbit defects, non-zero transverse fields are experi-
enced along the closed orbit, they excite “imperfection”, aka “integer”, depolarizing
resonances, located at
. Gγ R = k, k∈N

In the case that the periodicity of the orbit is that of the lattice, P, the sole imperfection
resonances, located at.Gγ R = k P, are excited. The strongest imperfection resonances
are located at [32, Sect. 3.II]
. Gγ R = k m P

Spin precession axis. Resonance width

Consider the spin vector
.S(θ ) = (Sη , Sξ , S y )
264 8 Weak Focusing Synchrotron

of a particle, in the laboratory frame, with .θ the orbital angle around the accelerator.
Introduce the projection .s(θ ) of .S in the bend plane

s(θ ) = Sη (θ ) + j Sξ (θ )
. (and S y2 = 1 − s 2 ) (8.47)

In the case of a stationary solution of the spin motion, viz. stationary spin preces-
sion axis around the ring (Fig. 8.21) [31, Sect. 3.6.1], .s satisfies [31] (Fig. 8.20)

1
.s2 = (8.48)
/\2
1+
|E R |2

with./\ = Gγ − Gγ R the distance to the resonance; thus the resonance width appears
to be a measure of its strength. The quantity of interest is the angle, .φ, of the spin
precession direction to the vertical axis.

/
Fig. 8.20 A graph of .s(/\) = 1 − S y2 (/\). .s = 1 on the resonance (./\ = 0), the spin vector lies in
√
the bend plane. .s = 1/2 at distance ./\ = ± 3E R from .Gγ R , the spin vector is at .30◦ to the . y axis

Fig. 8.21 Near an integer

resonance, at any azimuth .θ
around the ring spins .S(m)
(.m is the turn number, .S(m)
started vertical, here)
precess
/ at frequency
.ω = /\2 + |E R |2 around a
stationary axis .n0 (θ), whose
orientation varies along the
ring. .n0 is aligned along .S,
average of .S(m) over turns
8.2 Basic Concepts and Formulæ 265

Fig. 8.22 Dependence of polarization on the distance to the resonance. For instance.|S y | = 0.99, 1%
depolarization at./\ = ±7|E R |.. S y = 0, full depolarization on the resonance (./\ = 0), the precession
axis lies in the bend plane

It is given by (Fig. 8.22)

/\2 /|E R |2
. cos2 φ(/\) ≡ S y2 (/\) = 1 − s 2 = (8.49)
1 + /\2 /|E R |2

On the resonance, with ./\ = 0, the spin precession axis lies in the bend plane:
φ = ±π/2. A depolarization by 1% (|. S y | = 0.99) corresponds to a distance to the
.
resonance ./\ = 7|E R |, spin precession axis at an angle .φ = acos(0.99) = 8◦ from
the vertical.
Conversely,
/\2 S y2
. = (8.50)
|E R |2 1 − S y2

The precession axis is common to all spins, while . S y is a measure of the polarization
along the vertical axis,
N+ − N−
. Sy =
N+ + N−

where. N + and. N − denote the number of particles in spin states. 21 and.− 12 respectively.
Things complicate a little in the vicinity of an intrinsic resonance [31, Sect. 3.6.2],
the precession axis is not stationary, spins precess around it while it precesses itself
around the vertical, Fig. 8.23.
266 8 Weak Focusing Synchrotron

Fig. 8.23 Near an intrinsic

resonance, spins .S(m)
precess at frequency .ω
around an axis .n, which itself
precesses around the vertical
axis at frequency .Gγ

Resonance crossing
Crossing an isolated resonance (Figs. 8.19 and 8.24) polarization is affected accord-
ing to the Froissart-Stora law [34], [31, Sect. 2.3.6],

π |E R |2
= 2e− 2
Pf
. α −1 (8.51)
Pi

from a value . Pi upstream to an asymptotic value . P f downstream of the resonance.

In this expression
dγ 1 /\E
.α = G = (8.52)
dθ 2π M
is the crossing speed for an energy gain ./\E per turn.
Spin motion through weak resonances
Depolarizing resonances are weak up to several GeV in a weak focusing synchrotron
because the radial and/or longitudinal fields are weak. Thus assume . Sy,f ≈ Sy,i , with
. Sy,f and . Sy,i the asymptotic vertical spin component values respectively upstream
and downstream of the resonance. With the origin of the orbital angle taken at the
resonance (Fig. 8.24), and introducing the Fresnel integrals [31]
{ x (π ) { x (π )
.C(x) = cos t 2 dt, S(x) = sin t 2 dt
0 2 0 2
8.2 Basic Concepts and Formulæ 267

Fig. 8.24 Vertical component of spin motion . S y (θ) through a weak depolarizing resonance
(Eq. 8.53). The vertical line is at the location of the resonance, which coincides with the origin
of the orbital angle

the polarization satisfies

( )2 {[ / )]2 [ ( ( / )]2 }
Sy (θ) π |E R |2 1α 1 α
if θ < 0 : =1− −θ 2 −C+ 2 − S −θ
Sy,i α π π
. ( )2 {[ ( / )]2 [ ( / )]2 }
Sy (θ) π |E R |2 1 α 1 α
if θ > 0 : =1− +C θ + 2 +S θ
Sy,i α 2 π π
(8.53)
In the asymptotic limit,
S y (θ ) θ−→∞ π
. −→ 1 − |E R |2 (8.54)
Sy,i α

which agrees with a Taylor development of Froissart-Stora formula, Eq. 8.51, to first
order in .|E R |2 /α. This approximation holds in the limit that higher order terms can
be neglected.
268 8 Weak Focusing Synchrotron

8.3 Exercises

8.1 Construct SATURNE 1 Synchrotron. Spin Resonances

Solution 8.1.
In this exercise, the weak focusing 3 GeV synchrotron SATURNE 1 (Fig. 8.1) is
modeled. Spin resonances in a weak dipole gradient lattice are observed.
(a) Construct a model of SATURNE 1 90.◦ cell dipole in the hard-edge model,
using DIPOLE. Use the parameters given in Table 8.1, and Fig. 8.25 as a guidance.
For beam monitoring purposes, split the dipole in two 45.◦ halves. It is judicious to
take RM .= 841.93 cm in DIPOLE, as this is the reference radius for the definition
of the radial index. Take an integration step size in centimeter range—small enough
to ensure numerical convergence, as large as doable for faster multiturn raytracing.
Validate the model by producing the .6 × 6 transport matrix of the cell dipole
(MATRIX[IFOC=0] can be used for that, with OBJET[KOBJ=5] to define a proper
set of paraxial initial coordinates) and checking against theory (Sect. 14.1, Eq. 14.6).
(b) Construct a model of SATURNE 1 cell, with origin at the center of the drift.
Find the closed orbit, that particular trajectory which has all its coordinates zero in
the drifts: use DIPOLE[KPOS] to cancel the closed orbit coordinates at DIPOLE
ends. While there, check the expected value of the dispersion (Eq. 8.26) and of
the momentum compaction (Eq. 8.28), from the raytracing of a chromatic closed
orbit—i.e., the orbit of an off-momentum particle. Plot these two orbits (on- and
off-momentum), over a complete turn around the ring, on a common graph.

Table 8.1 Parameters of SATURNE 1 weak focusing synchrotron [35]. .ρ0 denotes the reference
bending radius in the dipole; the reference orbit, field index, wave numbers, etc., are taken along
that radius
Orbit length, .C cm 6890
Average radius, . R = C/2π cm 1096.58
Drift length, .2l cm 400
Magnetic radius, .ρ0 cm 841.93
. R/ρ0 = 1 + k 1.30246
Field index .n, nominal 0.6
Wave numbers .νx , .ν y , nominal 0.72, 0.89
Stability limit .0.5 < n < 0.757

Injection energy (proton) MeV 3.6

Field at injection kG 0.326
Top energy GeV 2.94
Field at top energy, . Bmax kG 14.9
. Ḃ kG/s 18
Synchronous energy gain keV/turn 1.160
RF harmonic 2
8.3 Exercises 269

Fig. 8.25 A schematic

layout of SATURNE 1, a
.2π/4 axial symmetry
structure, comprised of 4
radial field index 90.◦ dipoles
and 4 drift spaces. The cell in
the simulation exercises is
taken as a .π/2 quadrant:
half-drift / 90.◦ -dipole / half-
drift

Compute the cell periodic optical functions and tunes, using either
OBJET[KOBJ=5] and MATRIX[IORD=1,IFOC=11], or TWISS, or
OBJET[KOBJ=6] and MATRIX[IORD=2,IFOC=11]; check their values against
theory. Check consistency with previous dispersion function and momentum com-
paction outcomes.
Move the origin of the lattice at a different azimuth .s along the cell: verify that,
while the transport matrix depends on the origin, its trace does not.
Produce a graph of the optical functions (betatron functions and dispersion) along
the cell. Check the expected average values of the betatron functions (Eq. 8.20).
Produce a scan of the tunes over the field index range .0.5 ≤ n ≤ 0.757. REBE-
LOTE[IOPT=1] can be used to repeatedly change .n over that range. Superimpose
the theoretical curves .νx (n), .ν y (n).
(c) Justify considering the betatron oscillation as sinusoidal, namely,

. y(θ ) = A cos(ν y θ + φ)
{
wherein .θ = s/R, . R = ds/2π .
(d) Launch a few particles evenly distributed on a common paraxial horizontal
Courant-Snyder invariant, vertical motion taken null (OBJET[KOBJ=8] can be used),
for a single pass through the cell. Store particle data along the cell in zgoubi.plt, using
DIPOLE[IL=2] and DRIFT[split,N=20,IL=2]. Use these to generate a graph of the
beam envelopes.
270 8 Weak Focusing Synchrotron

Using Eq. 8.22 compare with the results obtained in (b). Find the minimum
and maximum values of the betatron functions, and their azimuth .s(min[βx ]),
.s(max[β x ]). Check the latter against theory.
Repeat for the vertical motion, taking .εx = 0, .ε y paraxial.
Repeat, using, instead of several particles on a common invariant, a single particle
traced over a few tens of turns.
(e) Produce an acceleration cycle from 3.6 MeV to 3 GeV, for a few particles
launched on a common .10−4 π m initial invariant in each plane. Ignore synchrotron
motion (CAVITE[IOPT=3] can be used in that case). Take a peak voltage .V̂ =
200 kV (for faster raytracing—unrealistic though, as it would result in prohibitive . Ḃ
(Eq. 8.29)) and synchronous phase .φs = 150◦ (justify .φs > π/2).
Check the betatron damping over the acceleration range: compare with theory
(Eq. 8.31).
How close to symplectic the numerical integration is (it is by definition not sym-
plectic, being a truncated Taylor series method [36, Eq. 1.2.4]), depends on the inte-
gration step size, and on the size of the flying mesh in the DIPOLE[IORDRE,Resol]
method [36, Fig. 20]; check a possible departure of the betatron damping from theory
as a function of these parameters.
Produce a graph of the horizontal and vertical wave number values over the accel-
eration cycle.
(f) Some spin motion, now. Adding SPNTRK at the beginning of the sequence
used in (e) will ensure spin tracking.
Based on the input data file worked out for question (d), simulate the acceleration
of a single particle, through the intrinsic resonance .Gγ R = 4 − ν y , from a distance
of a few times the resonance strength upstream (this requires determining BORO
value under OBJET) to a distance of a few times the resonance strength downstream
of the resonance, at an acceleration rate of 10 kV/turn.
OBJET[KOBJ=8] can be used to allow to easily define an initial invariant value.
Start with spin vertical. On a common graph, plot . S y (tur n) for a few differ-
ent values of the vertical betatron invariant (the horizontal invariant value does not
matter—explain that statement, it can be taken zero). Derive the resonance strength
from this tracking, check against theory.
Repeat, for different crossing speeds.
Push the tracking beyond .Gγ = 2 × 4 + ν y : verify that the sole systematic reso-
nances .Gγ = integer × P ± ν y are excited—with . P = 4 the periodicity of the ring.
Break the 4-periodicity of the lattice by perturbing the index in one of the 4 dipoles
(say, by 10%), verify that all resonances .Gγ = integer ± ν y are now excited.
(g) Consider a case of weak resonance crossing, single particle (i.e., a case where
. P f /Pi ≈ 1, taken from (f); crossing speed may be increased, or particle invariant
decreased if needed), show that it satisfies Eq. 8.53. Match its turn-by-turn tracking
data to Eq. 8.53 so to get the vertical betatron tune .ν y , the location of the resonance
. GγR , and its strength.
8.3 Exercises 271

(h) Stationary spin motion (i.e. at fixed energy) is considered in this question. Track
a few particles with distances from the resonance./\ = Gγ − Gγ R = Gγ − (4 − ν y )
evenly spanning the interval ./\ ∈ [0, 7 × E R ].
Produce on a common graph the spin motion . S y (tur n) for these particles, as
observed at some azimuth along the ring. ( )
Produce a graph of the average over turns, . S y |turn (/\) (as in Fig. 8.22). Produce
the vertical betatron tune .ν( y , )the location of the resonance .GγR , and its strength .E R ,
obtained from a match of . S y |turn (/\) to (Eq. 8.49)

|/\|
(S y )(/\) = /
.
|E R |2 + /\2

(i) Track a 200-particle 6-D bunch, with Gaussian transverse densities .εx,y a few
.µm, and Gaussian .δp/ p with .σδp/ p = 10−4 . Produce a graph of the average value of
the vertical spin component . S y over a 200 particle set, as a function of .Gγ , across the
. Gγ R = 4 − ν y resonance. Indicate on that graph the location of the resonant . Gγ R
values.
Perform this resonance crossing for five different values of the particle invariant:
.ε y /π = 2, 10, 20, 40, 200 µm. Compute. P f /Pi in each case, check the dependence
on .ε y against theory.
Compute the resonance strength, .ε y , from this tracking.
Re-do this crossing simulation for a different crossing speed (take for instance
. V̂ = 10 kV) and a couple of vertical invariant values, compute . P f /Pi so obtained.

Check the crossing speed dependence of . P f /Pi against theory.

8.2 Construct the ZGS Synchrotron. Spin Resonances

Solution 8.2.
In this exercise, the ZGS 12 GeV synchrotron is modeled. Spin resonances in a zero-
gradient, wedge focusing synchrotron are addressed.
A photo taken in the ZGS tunnel is given in Fig. 8.4; a schematic layout of the ring
is shown in Fig. 8.26, and a sketch of the double dipole cell in Fig. 8.27. Table 8.2
details the parameters of the synchrotron resorted to in these simulations.
(a) Construct a model of ZGS 45.◦ cell dipole in the hard-edge model, using
DIPOLE. Use the parameters given in Table 8.2, and Figs. 8.26 and 8.27 as a guidance.
For beam monitoring purposes, split the dipole in two 22.5.◦ halves. Take the closed
orbit radius as the reference RM .= 2076 cm in DIPOLE: it will be assumed that the
orbit is the same at all energies.7 Take an integration step size in centimeter range—
small enough to ensure numerical convergence, as large as doable for fast multiturn
raytracing.
Validate the model by producing the .6 × 6 transport matrices of both dipoles
(MATRIX[IFOC=0] can be used for that, with OBJET[KOBJ=5] to define a proper
set of paraxial initial coordinates) and checking against theory (Sect. 14.1, Eq. 14.6).

7 Note that in reality the reference orbit in ZGS moved outward during acceleration [37].
272 8 Weak Focusing Synchrotron

Fig. 8.26 A schematic layout of the ZGS [33], a .π -periodic structure, comprised of 8 zero-index
dipoles, 4 long and 4 short straight sections

Fig. 8.27 A sketch of ZGS

cell layout. In defining the
entrance and exit faces
(EFBs) of the magnet, beam
goes from left to right.
Wedge angles at the long
straight sections (.ε1 ) and at
the short straight sections
(.ε2 ) are different
8.3 Exercises 273

Table 8.2 Parameters of the ZGS weak focusing synchrotron after Refs. [37, 38] [33, pp. 288–294,
p. 716] (2nd column, when they are known) and in the present simplified model and numerical
simulations (3rd column). Note that the actual orbit moves during ZGS acceleration cycle, tunes
change as well—this is not taken into account in the present modeling, for simplicity
From Refs. [37, 38] Simplified model
Injection energy MeV 50
Top energy GeV 12.5
.Gγ span 1.888387–25.67781
Length of central orbit m 171.8 170.90457
Length of straight m 41.45 40.44
sections, total
Lattice
Wave numbers .νx ; ν y 0.82; 0.79 0.849; 0.771
Max. .βx ; β y m 32.5; 37.1
Magnet
Length m 16.3 .
16.30486
(magnetic)
Magnetic radius m 21.716 20.76
Field min.; max. kG 0.482; 21.5 0.4986; 21.54
Field index 0
Yoke angular extent deg 43.02590 45
Wedge angle deg .≈10 13 and 8
RF
Rev. frequency MHz 0.55–1.75 0.551–1.751
RF harmonic 8
h=.ωrf /ωrev
Peak voltage kV 20 200
B-dot, nominal/max. T/s 2.15/2.6
Energy gain, keV/turn 8.3/10 100
nominal/max.
Synchronous phase, deg 150
nominal
Beam
.εx ; .ε y (at injection) .π µm 25; 150
Momentum spread, .3× 10−4
rms
Polarization at % .>75 100
injection
274 8 Weak Focusing Synchrotron

Add fringe fields in DIPOLE[.λ,.C0 − C5 ], the rest of the exercise will use that
model. Take fringe field extent and coefficient values

.λ = 60 cm, C0 = 0.1455, C1 = 2.2670, C2 = −0.6395, C3 = 1.1558, C4 = C5 = 0

(8.55)

(.C0 − C5 determine the shape of the field fall-off, they have been computed from a
typical measured field profile . B(s)).
(b) Construct a model of ZGS cell accounting for dipole fringe fields, with origin
at the center of the long drift. In doing so, use DIPOLE[KPOS] to cancel the closed
orbit coordinates at DIPOLE ends.
Compute the periodic optical functions at cell ends, and cell tunes, using
MATRIX[IORD=1,IFOC=11] (or OBJET[KOBJ=6] and MATRIX[IORD=2,
IFOC=11]); check their values against theory.
Move the origin at the location (azimuth .s along the cell) of the betatron functions
extrema: verify that, while the transport matrix depends on the origin, its trace does
not. Verify that the local betatron function extrema, and the dispersion function, have
the expected values.
Produce a graph of the optical functions (betatron functions and dispersion) along
the cell.
(c) Additional verifications regarding the model.
Produce a graph of the field B(s)
– along the on-momentum closed orbit, and along off-momentum chromatic
closed orbits, across a cell;
– along orbits at large horizontal excursion;
– along orbits at large vertical excursion.
For all these cases, verify qualitatively, from the graphs, that . B(s) appears as
expected.
(d) Justify considering the betatron oscillation as sinusoidal, namely,

. y(θ ) = A cos(ν y θ + φ)
{
wherein .θ = s/R, . R = ds/2π .
(e) Produce an acceleration cycle from 50 MeV to ∼17 GeV about, for a few
particles launched on a common .10−5 π m vertical initial invariant, with small hori-
zontal invariant. Ignore synchrotron motion (CAVITE[IOPT=3] can be used in that
case). Take a peak voltage .V̂ = 200 kV (this is unrealistic but yields 10 times faster
computing than the actual .V̂ = 20 kV, Table 8.2) and synchronous phase .φs = 150◦
(justify .φs > π/2). Add spin, using SPNTRK, in view of the next question, (f).
Check the accuracy of the betatron damping over the acceleration range, compared
to theory. How close to symplectic the numerical integration is (it is by definition
not symplectic [29, Eq. 1.2.4]), depends on the integration step size, and on the size
of the flying mesh in the DIPOLE method [36, Fig. 20]; check a possible departure
of the betatron damping from theory as a function of these parameters.
8.4 Solutions of Exercises of This Chapter: Weak Focusing Synchrotron 275

Produce a graph of the evolution of the horizontal and vertical wave numbers
during the acceleration cycle.
(f) Using the raytracing material developed in (e): produce a graph of the vertical
spin component of a few particles, and the average value over the 200 particle bunch,
as a function of .Gγ . Indicate on that graph the location of the resonant .Gγ R values.
(g) Based on the simulation file used in (f), simulate the acceleration of a sin-
gle particle, through one particular intrinsic resonance, from a few thousand turns
upstream to a few thousand turns downstream.
Perform this resonance crossing for different values of the particle invariant. Deter-
mine the dependence of final/initial vertical spin component value, on the invariant
value; check against theory.
Re-do this crossing simulation for a different crossing speed. Check the crossing
speed dependence of final/initial vertical spin component so obtained, against theory.
(h) Introduce a vertical orbit defect in the ZGS ring.
Find the closed orbit.
Accelerate a particle launched on that closed orbit, from 50 MeV to ∼17 GeV
about, produce a graph of the vertical spin component.
Select one particular resonance, reproduce the two methods of (g) to check the
location of the resonance at .Gγ R = integer, and to find its strength.

8.3 Visible SR from GEC 70 MeV Synchrotron

Solution 8.3.
Produce the electric field impulse radiated by a 70 MeV electron in GEC synchrotron,
as observed at a vertical elevation .ψ /= 0 in a plane tangent to the orbit. MULTIPOL
can be used to simulate a trajectory arc of a few .1/γ deviation. Set IL=2 to log
stepwise particle coordinates in zgoubi.plt.
Produce the spectral brightness of the radiation in that direction.
Zpop menu 8/16 can be used for these two questions. An alternative is to program
the equations of concern (Eq. 8.36 et seqs.) in an interface, in python for instance.
The 70 MeV orbit radius in GEC synchrotron is 29.2 cm.

8.4 Solutions of Exercises of This Chapter: Weak Focusing

Synchrotron

8.1 Construct SATURNE 1 Synchrotron. Spin Resonances

Figure 8.1 displays a photo of SATURNE 1 synchrotron. A schematic layout of the
ring and 90.◦ cell is given in Fig. 8.25. This figure and Table 8.1 will be referred to
in building SATURNE 1 ring in the following.
276 8 Weak Focusing Synchrotron

Fig. 8.28 A representation of the data that define a dipole magnet, using DIPOLE [36]

(a) A model of SATURNE 1 synchrotron.

DIPOLE is used to simulate the 90.◦ cell dipole, data are set for a hard-edge model
in this exercise (for a DIPOLE model including fringe fields, refer to the ZGS case,
Exercise 8.2).
Some guidance regarding DIPOLE data, referring to Fig. 8.28:
• DIPOLE is defined in a cylindrical coordinate system.
• . AT is given the value of the bending sector extent: . AT = 90◦ . The dipole EFBs
coincide with DIPOLE entrance and exit boundaries.
• . R M is given the curvature radius value, . R M = Bρ/B = 0.274426548 [T m]/
0.03259493 [T] = 8.4193 m, as it fits the geometry of the optical axis around the
ring. The field value matches the reference rigidity under OBJET, these are the
injection energy values, 3.6 MeV, proton.
• ACENT=45.◦ is the reference azimuth, for the positioning of the entrance and exit
EFBs. It is taken half-way of the . AT range, an arbitrary choice.
• KPOS=2 allows cancelling the coordinates of particle 1 (considered here as the
reference trajectory, coinciding with the optical axis around the ring) at entrance
and exit of DIPOLE:
– The entrance and exit radii in and out of the . AT sector for a particle on the
closed orbit (i.e., a particle traveling along the design optical axis) are . R E =
RS = R M.
8.4 Solutions of Exercises of This Chapter: Weak Focusing Synchrotron 277

– The angle TE identifies with the closed orbit angle at the entrance EFB: TE=0,
the closed orbit is normal to the EFB. Same for TS at exit EFB.

Simulation of a 90.◦ sector in the hard edge model is given in Table 8.3; note that the
sector has been split in two 45.◦ halves, this is in order to allow a possible insertion of
a beam monitor, so requiring . AT = 45◦ , .ω+ = −ω− = 22.5◦ . FAISCEAU located
next to DIPOLE indicates that a trajectory entering DIPOLE at radius . R = R M,
normally to the EFB (thus, .Y0 = 0 and .T0 = 0 in OBJET) exits with .Y = 0 and
. T = 0. Data validation at this stage can be performed by comparing DIPOLE’s
transport matrix computed with MATRIX (Table 8.4), and theoretical expectations
(Sect. 14.1, Eq. 14.6):
⎛ ⎞
α=π/2, 0.545794 11.15444 0 0 0 9.560222
ρ=8.4193 ⎜ 0.062944 0.545794 0 0 0 1.324865 ⎟
⎜ ⎟
n=0.6 ⎜ 0 0 0.346711 10.19506 0 0 ⎟
.[Tij ] ⎜ ⎟ (8.56)
= ⎜ 0 0 −0.086295 0.346711 0 0 ⎟
⎜ ⎟
(Eq. 14.6) ⎝ 1.324865 9.560222 0 0 1 5.17640 ⎠
0 0 0 0 0 1

Introducing fringe fields

SATURNE ring simulations which follow use the hard edge model, however, it
might be desired at this point to add DIPOLE fringe fields. The following changes
are needed in that case:

• The bending sector is 90.◦ , however the field region extent . AT has to encom-
pass the fringe fields, at both ends of the 90.◦ sector. A 5.◦ extension is taken
(namely, . AC E N T − ω+ = AT − AC E N T + ω− = 5◦ ), for a total . AT = 100◦
which allows . R M × tan(AC E N T − ω+ ) ≈ 74 cm; this large extension ensures
absence of truncation of the fringe fields at the . AT sector boundaries, over the all
radial excursion of the beam.
• ACENT .= 50◦ is the reference azimuth (an arbitrary value; taken half-way of the
. AT range for convenience), for the positioning of the entrance and exit EFBs.
• The entrance radius in the . AT sector is . R E = R M/ cos(AC E N T − ω+ ) =
R M/ cos(5◦ ), with .ω+ = 45◦ the positioning of the entrance EFB with respect
to ACENT. And similarly for the positioning of the exit reference frame, . RS =
R M/ cos(AT − (AC E N T − ω− )) = R M/ cos(5◦ ) with .ω− = −45◦ the posi-
tioning of the exit EFB. Note that .ω+ − ω− = 90◦ , the value of the sector angle.
• The entrance angle TE identifies with the angular increase of the sector: TE=5.◦ .
And similarly for the positioning of exit frame, 5.◦ downstream of the exit EFB,
thus TS .= 5◦ .
• Negative drifts with equal lengths
.R M × tan(AC E N T − ω+ ) = R M × tan(AT − (AC E N T − ω− )) = 0.7366545469 cm
need to be added upstream and downstream of DIPOLE, to account for the optical
axis additional length over the 5.◦ angular extent.
278 8 Weak Focusing Synchrotron

Table 8.3 Simulation input data file: SATURNE 1 90.◦ DIPOLE is split into a pair of adjacent 45.◦
sectors in the hard edge model. FAISTORE or (here) FAISCEAU is inserted between the latter two,
for beam monitoring. The reference optical axis has equal entrance (RE) and exit (RS) positions,
and null angles (TE and TS), it coincides with the arc of radius. R = R M inside the sector. This input
data file is named SatI._DIP.inc and defines the cell sequence segment S._SatI._DIP to E._SatI._DIP,
for INCLUDE statements in subsequent exercises. The present file equipped with OBJET[KOBJ=5]
and MATRIX[IFOC=0] computes the dipole transport matrix

(b) SATURNE 1 cell.

A cell with origin in the middle of the drift is given Table 8.5, it INCLUDEs the
split dipole, with a pair of 2 m half-drifts added at both ends (Fig. 8.25).
Closed orbit; chromatic closed orbit
The on-momentum closed orbit is set to zero along the drifts (see coordinates in
Table 8.4: .Y0 = Y = 0, .T0 = T = 0) thanks to DIPOLE[KPOS=2, . R E = RS =
R M, .T E = T S = 0] (Table 8.3).
8.4 Solutions of Exercises of This Chapter: Weak Focusing Synchrotron 279

Table 8.4 Outcomes of the DIPOLE simulation of Table 8.3

The radial coordinate of an off-momentum chromatic orbit can be estimated from

the dispersion, Eq. 8.26, namely,

ρ0 δp 841.93
Y =
. δ = 10−4 ≈ 0.21048 cm
1−n p 1 − (−0.6)

whereas the orbit angle is zero, around the ring (on- and off-momentum closed orbits
are parallel to the optical axis).
Besides,
– computation of an accurate value of .Yδ is performed adding FIT at the end of
the cell;
280 8 Weak Focusing Synchrotron

Table 8.5 Simulation input data file: SATURNE 1 cell, assembled by INCLUDE-ing DIPOLE
taken from Table 8.3 together with two half-drifts. This input data file is named SatI_cell.inc
and defines the SATURNE 1 cell sequence segment S_SatI_cell to E_SatI_cell, for INCLUDE
statements in subsequent exercises

– in order to raytrace three particles, respectively on-momentum and at .δp/ p =

±10−4 , OBJET[KOBJ=2] is used;
– in order to raytrace around the ring, for the purpose of plotting the closed orbit
coordinates, a 4-cell sequence follows the FIT procedure.
This results in the input data file given in Table 8.6. Running this file produces the
following coordinates as per the FIT procedure (an excerpt from zgoubi.res execution
listing):

The local coordinates.Y ,.T and initial coordinates.Y0 ,.T0 (as defined under OBJET)
are identical to better than .5 µm, .0.5 µrad, respectively, confirming the periodicity
of these chromatic trajectories. Orbit coordinates around the ring are displayed in
Fig. 8.29.
8.4 Solutions of Exercises of This Chapter: Weak Focusing Synchrotron 281

Table 8.6 Simulation input data file: first find the three periodic orbits at .δp/ p = 0, ±10−4 ,
through a cell; once FIT is done, complete a 4-cell turn

Fig. 8.29 Radial coordinate of the orbits around the ring, on-momentum, and for .dp/ p = ±10−3 .
A graph obtained using zpop: menu 7; 1/1 to open zgoubi.plt; 2/[6,2] for .Y versus distance .s; 7 to
plot. A gnuplot script for a similar graph is given in Table 8.6
282 8 Weak Focusing Synchrotron

Table 8.7 Results obtained running the simulation input data file of Table 8.5, SATURNE 1 cell—
an excerpt from zgoubi.res execution listing

Table 8.8 The header part of zgoubi.TWISS.out listing resulting from the SATURNE 1 cell sim-
ulation of Table 8.5. The ring wave numbers are 4 times the present cell values Q1, Q2. Optical
functions (betatron function and derivative, orbit, phase advance, etc.) along the optical sequence
follow this header in zgoubi.TWISS.out

Lattice parameters
The TWISS command down the sequence (Table 8.5) produces the periodic beam
matrix results shown in Table 8.7; MATRIX[IFOC=11] would, as well. It also pro-
duces a zgoubi.TWISS.out file which details the optical functions along the sequence
(at the downstream end of the optical elements). The header of that file details the
optical parameters of the structure (Table 8.8).
Moving the origin of the cell
The origin of the sequence can be moved by placing both drifts at an end of DIPOLE.
It can also be taken in the middle of DIPOLE, as the latter has been split. An input data
sequence (with INCLUDEs expanded) is provided at the top of the execution listing
zgoubi.res, it can be used to copy-paste pieces around. It can then be checked that
8.4 Solutions of Exercises of This Chapter: Weak Focusing Synchrotron 283

betatron tunes, chromaticities, momentum compaction (Table 8.7) do not change,

whereas the beam matrix does.
Optical functions along the cell
They are computed by transporting the beam matrix, from the origin. A Fortran pro-
gram available in zgoubi sourceforge package toolbox, betaFromPlt [36], performs
this computation in the following way: OBJET[KOBJ=5.1] provides the initial beta
function values (determined in the previous question); IL=2 under DIPOLE logs
stepwise particle data in zgoubi.plt; ‘split 10 2’ added under DRIFT does it, too [36,
DRIFT]. The program betaFromPlt computes the transport matrix .Tstepi from the
origin of the sequence (at OBJET) to the considered .stepi along the sequence, using
particle coordinates read in zgoubi.plt—a similar [computation ] to what MATRIX
β −α
does [36, MATRIX Sect.]. The beam matrix .σ = is then transported,
−α γ
from the origin to .stepi , using (Eq. 14.46)

σ
. stepi = Tstepi σorigin T̃stepi

The result is displayed in Fig. 8.30.

Tune scan
A simulation is given in Table 8.9, derived from Table 8.5: MATRIX[IFOC=11] has
been substituted to TWISS, a REBELOTE do loop repeatedly changes .n. A graph
of the scan is given in Fig. 8.31, a few values are detailed in Table 8.10.

Fig. 8.30 Optical functions along SATURNE 1 cell. They are obtained from the transport of the
beta functions, from the origin (at OBJET), using transport matrices computed from step-by-step
particle coordinates stored in zgoubi.plt
284 8 Weak Focusing Synchrotron

Table 8.9 Simulation input data file: tune scan, using REBELOTE to repeatedly change DIPOLE
field index .n. Beam matrix and wave numbers are computed by MATRIX, from the coordinates of
the 13 particle sample generated by OBJET[KOBJ=5]

(c) Sinusoidal approximation of the betatron motion.

The approximation
. y(θ ) = A cos(ν Z θ + φ)

is checked here considering the vertical motion (considering the horizontal motion
leads to similar conclusions). The value of the various parameters in that expression
are determined as follows:
– the particle raytraced for comparison is launched with an initial excursion
. Z 0 (θ = 0) = 5 cm. At the launch point (middle of the drift) the beam ellipse is
upright (see below), whereas phase space motion is clockwise, thus take
8.4 Solutions of Exercises of This Chapter: Weak Focusing Synchrotron 285

/ √
Fig. 8.31 A scan of the wave numbers, and of . νY2 + ν Z2 ≈ R/ρ0 = 1.141, in SATURNE 1 for
.0.5 ≤ n ≤ 0.757. Solid curves are from theoretical approximations (Eq. 8.23), markers are from
numerical simulations

Table 8.10 Dependence of wave numbers on index.n, from numerical raytracing (columns denoted
“ray-tr.”) and from theory
n .νY .ν Z
/ /
. (1 − n)
R R
ray-tr. ρ0 ray-tr. . n
ρ0
0.5 0.810353 0.806987 0.810353 0.806987
0.6 0.724125 0.721791 0.888583 0.884010
0.7 0.626561 0.625089 0.960806 0.954840
0.757 0.563635 0.562580 0.999804 0.992955

A = 5 cm and φ = 0
.

– the vertical betatron tune of the 4-cell ring is (Table 8.8)

ν = 4 × 0.222146 = 0.888284
. Z

{
– .θ = s/R and . R = ds/2π with (Table 8.8)

.2π R = circumference = 2π × 10.9658 = 68.9 m

Consistency with sinusoidal approximation is shown in Fig. 8.32.

286 8 Weak Focusing Synchrotron

Fig. 8.32 Vertical betatron motion, five turns around SATURNE 1 ring, from raytracing (modulated
oscillation), and sine approximation, superimposed

(d) Beam envelopes, periodic ellipses.

A few particles are launched through the cell with initial coordinates taken
on a common invariant (horizontal and/or vertical), using OBJET[KOBJ=8]. The
input data file is given in Table 8.11. The initial ellipse parameters (under OBJET)
are the periodic values .αY = α Z = 0, .βY = 14.426 m, .β Z = 11.411 m, found in
zgoubi.TWISS.out (Table 8.8). The envelopes so generated, and the quantities
.u (s)/ εu /π (Eq. 8.22), are displayed in Fig. 8.33. The extremum extremorum of
2

.u (s)/ εu /π comes out to be, respectively, .β̂Y = 15.7 m and .β̂ Z = 13.08 m, consis-
2

tent with earlier derivations (BETXMAX and BETYMAX values in Table 8.8 and
Fig. 8.30).
This raytracing also provides the coordinates of the particles on their common
upright invariant (Fig. 8.34)

u 2 /βu + βu u ,2 = εu /π
.

at start and at the end of the cell (.εu /π = 10−4 , here). This allows checking that the
initial ellipse parameters (under OBJET, Table 8.11) are effectively periodic values,
and that the raytracing went correctly, namely by observing that the initial and final
ellipses do superimpose.
(e) An acceleration cycle. Symplecticity checks.
Eleven particles are launched for a 30,000 turn acceleration at a rate of

./\W = q V̂ sin φs = 200 × sin 1500 = 100 keV/turn

8.4 Solutions of Exercises of This Chapter: Weak Focusing Synchrotron 287

Table 8.11 Simulation input data file: raytrace 60 particles across SATURNE 1 cell to generate
beam envelopes. Store particle data in zgoubi.plt, along DIPOLEs and split DRIFTs. The INCLUDE
file and segment are defined in Table 8.5

Fig. 8.33 Left: horizontal and vertical envelopes as generated by plotting the coordinates Y(s)
(greater excursion, red, along the drifts and dipole) or Z(s) (smaller excursion, blue) across the
SATURNE 1 cell, of 60 particles evenly distributed on a common .10−4 π m invariant, either hori-
zontal or vertical (while the other invariant is zero). Right: a plot of .Y 2 (s)/ εY /π and . Z 2 (s)/ ε Z /π ;
their extrema identify with .βY (s) and .β Z (s), respectively. Graphs obtained using zpop: menu 7;
1/1 to open zgoubi.plt; 2/[6,2] (or [6,4]) for .Y versus .s (or . Z versus .s); 7 to plot; option 3/14 to
raise Y (or Z) to the square and normalize to .εY,Z /π

(. E : 3.6 MeV → 3.0036 GeV), all evenly distributed on the same initial vertical
invariant
,2
. Z /β Z + β Z Z = ε Z /π
2
(8.57)

with .ε Z /π = 10−4 m, or, normalized, .βγ ε Z /π = 0.08768 × 10−4 m.

288 8 Weak Focusing Synchrotron

Fig. 8.34 Sixty particles evenly distributed on a common periodic invariant (either .εY = 10−4 π m
and .ε Z = 0, left graph, or the reverse, right graph) have been tracked through the cell. Initial and
final phase space coordinates are displayed in these graphs: the initial and final ellipses superimpose.
Optical function values given in the figures result from an rms match, of indifferently the initial or
final coordinates; they do agree with TWISS outcomes (Table 8.8). A graph obtained using zpop:
menu 7; 1/5 to open zgoubi.fai; 2/[2,3] (or [4,5]) for .T versus .Y (or . P vs. . Z ); 7 to plot

The simulation file is given in Table 8.12. CAVITE[IOPT=3] is used, it provides

an RF phase independent boost

./\W = q V̂ sin φs

as including synchrotron motion is not necessary here, even better, anticipating on

spin question (f) this ensures constant depolarizing resonance crossing speed, so
precluding any possibility of multiple crossing (it can be referred to [40] regarding
that effect).
Betatron damping
Figure 8.35 shows the damped vertical motion of the individual particles, over the
acceleration range, together with the initial and final distributions of the 11 parti-
cles on elliptical invariants. Departure from the matching ellipse at the end of the
acceleration cycle, 3 GeV (Eq. 8.57 with .ε Z /π = 1.0745 × 10−6 m), is marginal.
Degree of non-symplecticity of the numerical integration
The degree of non-symplecticity as a function of integration step size is illustrated
in Fig. 8.36. The initial motion is taken paraxial, vertical motion is considered as it
resorts to off-mid plane Taylor expansion of fields [36, DIPOLE Sect.], a stringent
test as the latter is expected to deteriorate further the non-symplecticity inherent
to the Lorentz equation integration method (a truncated Taylor series method [36,
Eq. 1.2.4]).
Evolution of the wave numbers
The Fortran tool tunesFromFai_iterate can be used to computes tunes as a function
of turn number or energy, it reads turn-by-turn particle data from zgoubi.fai and
computes a discrete Fourier transform over so many turns (a few tens, 100 here
8.4 Solutions of Exercises of This Chapter: Weak Focusing Synchrotron 289

Table 8.12 Simulation input data file: track 11 particles launched on the same vertical invariant.
The INCLUDE adds the SATURNE 1 cell four times, the latter is defined in Table 8.5

for instance), every so many turns (300, here) [41]. Typical results are displayed in
Fig. 8.37, tunes have the expected values: .νY = 0.7241, .ν Z = 0.8885. An accelera-
tion rate of 100 keV/turn has been taken (namely, .V̂ = 200 kV and still .φs = 150◦ ),
to save on computing time. SCALING with option NTIM=.−1 causes the magnet
field to strictly follow the momentum boost by CAVITE.
(f) Crossing an isolated intrinsic depolarizing resonance.
The simulation uses the input data file of Table 8.12, with the following changes:
• Under OBJET:
– 1st line, change the reference rigidity BORO for an initial .Gγ ≈ 2.95, upstream
of .Gγ R = 4 − ν Z ≈ 3.1;
– 3rd line, request a single particle (“1 1 1”, in lieu of 11, “1 11 1”);
290 8 Weak Focusing Synchrotron

Fig. 8.35 Left: damped vertical motion, from 3.6 MeV to 3.004 GeV in 30,000 turns. Right: the ini-
tial coordinates of the 11 particles (squares) are taken on a common invariant .ε Z (0) = 10−4 π m (at
3.6 MeV,.βγ = 0.0877, thus.βγ ε Z (0) = 8.77 × 10−6 π m ); the final coordinates after 30,000 turns
(crosses) appear to still be (with negligible departure) on a common invariant, of value.ε Z ( f inal) =
2.149 × 10−6 π m, or.βγ ε Z ( f inal) = 8.77 × 10−6 π m (at 3.004 GeV,.βγ = 4.08045), equal to the
initial value .βγ ε Z (0)

Fig. 8.36 Turn-by-turn evolution of the normalized invariant,.βγ ε Z (tur n)/βγ ε Z (0) (initial.ε Z (0)
taken paraxial), for integration step sizes 1, 2 and 4 cm

– 6th line, set the invariant .ε Z /π to the desired value, .εY /π value is indifferent.
Resulting OBJET:
8.4 Solutions of Exercises of This Chapter: Weak Focusing Synchrotron 291

Fig. 8.37 Horizontal fractional ring tune (left vertical axis), .νY ≈ 0.7241, and vertical ring tune
(right vertical axis), .ν Z ≈ 0.8885, as a function of turn number, over 30,000 turns (. E : 0.0036 →
3 GeV at a rate of 100 keV/turn)

• change the field value under DIPOLE (via SCALING for instance) consistently
with the new BORO value, so to maintain a curvature radius .ρ0 = B O R O/B =
8.4193 m (Table 8.1),
• under CAVITE, set the peak voltage to the required value,
• under REBELOTE, set the number of turns to an appropriate value: a total of
15,000, of which 8,000 about upstream of the resonance, is convenient for an
acceleration rate of 10 keV/turn.
Changing the particle invariant value
Particle spin motion through the isolated resonance for seven different invariant
values, .ε Z /π = 1, 2, 10, 20, 40, 80, 200 µm, observed at the beginning of the
optical sequence (FAISTORE[b_polarLand.fai] location, Table 8.12), is displayed
in Fig. 8.38.
The intrinsic resonance strength satisfies.|E R |2 = A ε Z , with. A a factor which char-
acterizes the lattice (see Sect.9.2.7, Eq. 9.53). On the other hand, from the Froissart-
Stora formula (Eq. 8.51) one gets
( ) ( )
2α 2 SZ,f ≈SZ,i α SZ,f
.|E R |2 = ln −→ 1− (8.58)
π 1 + SZ,f /SZ,i π SZ,i

with (.α, crossing ) speed (Eq. 8.52), a constant. Thus one expects to find
1 2
.
εZ
ln 1+SZ,f /SZ,i
constant.
Calculation of the resonance strength from . P f /Pi tracking outcomes, using
Eq. 8.58, requires the value of the crossing speed, which is
292 8 Weak Focusing Synchrotron

Fig. 8.38 Turn-by-turn spin motion through the isolated resonance .Gγ R = 4 − ν Z , for 7 different
values of the particle invariant from (top to bottom) .1 µm to .200 µm where full spin flip occurs.
A graph obtained using zpop: menu 7; 1/8 to open b_polarLand.fai; 2/[39,23] for . S Z versus turn;
7 to plot
( )
Table 8.13 Relationship between the invariant value .ε Z /π and the quantity .ln 2
1+SZ,f /SZ,i ∝
|E R |2 (Eq. 8.58). .V̂ = 20 kV, here, crossing speed .α = 1.696 × 10−6 (Eq. 8.59). . SZ,i = 1 always,
and . SZ,f (col. 2) is a rough estimate from Fig. 8.38. The rightmost column gives the resulting ratio
.|E R | /ε Z /π , essentially constant
2

|E R |2
(.×10−8 )
SZ,f
.ε Z /π (.µ m)
SZ,i ≡ SZ,f
2
. .ln .
1+SZ,f ε Z /π
1 0.89 0.024568 2.652645
2 0.795 0.046965 2.535451
10 0.17 0.232844 2.514034
20 .−0.35 0.488116 2.635115
40 .−0.78 0.958607 2.587537
80 .−0.975 1.903089 2.568474

1 /\E 1 20 × 103 × sin 30◦ [eV/turn]

α=
. = = 1.696 × 10−6 (8.59)
2π M 2π 938.27208 × 106 [eV]

Table 8.13, rightmost column, displays the ratio .|E R |2 /ε Z /π so obtained, essentially
constant as expected.
8.4 Solutions of Exercises of This Chapter: Weak Focusing Synchrotron 293
( )
Table 8.14 Relationship between the acceleration rate ./\E ∝ V̂ and the quantity .ln 1+SZ,f2 /SZ,i .
Normalized to.ε Z /π , their product (rightmost column) appears to be essentially constant, as expected

SZ,f
.ε Z /π (.µm) ≡ SZ,f × ln
2 V̂ 2
. V̂ (kV) . .ln .
SZ,i 1+SZ,f ε Z /π 1+SZ,f
1 10 .+0.79 0.048 0.482
10 10 .−0.33 0.475 0.475
20 10 .−0.78 0.959 0.479
1 20 .+0.89 0.025 0.49
2 20 .+0.795 0.047 0.47

Fig. 8.39 Resonance crossing in SATURNE 1, a turn-by-turn record of. S Z (Gγ ). Case of systematic
resonances .Gγ = 4k ± ν Z in a 4-period lattice (red), and of random resonances .Gγ = k ± ν Z in
a 1-periodic perturbed optics lattice (blue). A graph obtained using zpop: menu 7; 1/8 to open
b_polarLand.fai; 2/[59,23] for . S Z versus .Gγ ; 7 to plot

Changing the crossing speed

The crossing speed is reduced by a factor of 2, using .V̂ = 10 kV, and accordingly
the number of turns is doubled, to 30,000, the only modifications to the input data
simulation(file used in ) the previous question. Tracking results, Table 8.14, show that
.
V̂
ε Z /π
× ln 1+SZ,f /SZ,i is constant, as expected.
2

Systematic resonances, random resonances

A single-particle tracking is pushed beyond .Gγ = 8 + ν Z ≈ 8.89, 40,000 turns at
a rate of 100 kV/turn. The resulting . S Z (Gγ ), Fig. 8.39, shows that in a 4-periodic
lattice the sole systematic resonances are excited, whereas all resonances are excited
if the 4-periodicity is broken—here, by changing the index to .n = −0.66 in one
DIPOLE, the periodicity is 1.
294 8 Weak Focusing Synchrotron

(g) Spin motion across a weak depolarizing resonance.

The goal is to check numerical outcomes against the Fresnel integral model
(Eq. 8.53). A weak resonance is obtained using small amplitude vertical motion
and fast crossing.
A single particle is raytraced, in the following conditions:
– resonance to be crossed: .Gγ R = 4 − ν y ≈ 3.1115,
– acceleration: peak voltage .V̂ = 100 kV, synchronous phase .φs = 30◦ ,
– particle invariant .ε Z /π = 10−6 m.
The initial rigidity is taken a few hundred turns upstream of the resonance, namely,
. Bρref = 4.0880774 T m, 605.226550 MeV, . Gγ = 2.94931241, a distance to . Gγ R of
.4 − ν Z − 2.949312415 ≈ 0.16223. Tracking extends a few thousand turns beyond

. Gγ R so that . S Z reaches its asymptotic value, from which the resonance strength .|E R |
can be calculated, using Eq. 8.58.
The simulation file is given in Table 8.15. Note the new setting of the SCAL-
ING factor SCL: DIPOLE field was set for a curvature radius .ρ0 = 8.4193 m, given
a reference rigidity . Bρref ≡ B O R O = 0.274426548 Tm (Table 8.3). However the
reference rigidity is now changed to . Bρref = 4.0880774 T m, thus maintaining .ρ0
requires scaling the field in DIPOLE by .4.0880774/0.274426548 = 14.8968 at
turn 1: this is the new factor, . SC L = 14.8968, under SCALING (Table 8.15). Option
NT=.−1 under SCALING ensures that the scaling factor will automatically follow,
turn-by-turn, the rigidity boost by CAVITE so preserving constant curvature radius
.ρ0 = 8.4193 m.
The resulting turn-by-turn spin motion is displayed in Fig. 8.40. The Fresnel
integral model (Eq. 8.53) has been superimposed. Parameters in the latter are as
follows:
1 /\E 1 105 × sin 30◦ [eV/turn]
– crossing speed .α = = = 8.4812 × 10−6 ,
2π M 2π 938.27208 × 106 [eV]
– asymptotic . SZ,f = 0.999780, whereas initial . SZ,i = 1, thus (Eq. 8.58)

|E R |2 = 5.939 × 10−10
.

– orbital angle origin set at the location of .Gγ R , which is turn 1699.
(h) Stationary spin motion near a resonance.
The simulation input data file of Table 8.15 can be used for these fixed energy
trials, with some changes, as follows:
– OBJET[KOBJ=1] is used as it allows to define a set of particles with sampled
momentum offset, namely:
8.4 Solutions of Exercises of This Chapter: Weak Focusing Synchrotron 295

Table 8.15 Simulation input data file: track a particle launched on a vertical invariant .ε y /π =
10−6 m, with horizontal motion indifferent, taken zero here. The INCLUDE adds the SATURNE 1
cell four times, the latter is defined in Table 8.5

Fig. 8.40 Turn-by-turn spin motion through the isolated resonance .Gγ R = 4 − ν Z , case of weak
resonance strength. Modulated curve (blue): from raytracing. Smooth curve (black): Fresnel integral
model
296 8 Weak Focusing Synchrotron

Fig. 8.41 Turn-by-turn value of the vertical component of spins precessing at fixed energy in
SATURNE 1, observed at the beginning of the sequence, where spins start vertical (. S Z = 1). The
greater (respectively smaller) the distance to the resonance, the closer the precession axis is to
the vertical
/ (resp., to the bend plane), and the greater (resp. smaller) the oscillation frequency
.ω = /\2 + |E R |2

– with BORO changed, closer to .Gγ R = 4 − ν y ≈ 3.1115, DIPOLE field needs

to be set to 5.27284 kG,
– a number of turns REBELOTE[IPASS.≈a few thousand] results in at least half
an oscillation of . S Z (tur n) (the precession frequency increases with the distance to
the resonance, with a minimum of .ω = |E R | on the resonance [43, Fig. 3.4]), which
is convenient for determining .(S Z ).
Figure 8.41 displays the turn-by-turn evolution of the vertical component of the
spins as they precess around the eigenvector .n [43, Sect. 3.6.2, Fig. 3.3].
A quick, and accurate enough, approximation to the vertical component of the pre-
cession axis is .(S Z )|period = 21 {min [S Z (θ )] + max [S Z (θ )]}, it yields the .(S Z ) (/\)
graph of Fig. 8.42.
A match of the .(S Z ) values by (Eq. 8.49)

|/\|
|S Z (/\)| = /
.
/\2 + |E R |2

given .Gγ R = 4 − ν Z , yield vertical tune and resonance strength values, respectively,

ν = 0.88845
. Z and |E R | = 2.77 × 10−4

The vertical tune .ν Z is fairly consistent with earlier results, and .|E R | = 2.77 × 10−4
for .ε Z /π = 79 × 10−6 m also is with .|E R | = 2.44 × 10−5 for .ε Z /π = 10−6 m in the
previous question (h). The difference deserves further inspection, this is left to the
reader.
8.4 Solutions of Exercises of This Chapter: Weak Focusing Synchrotron 297

Fig. 8.42 Vertical component (absolute value) of the spin precession axis as a function of .Gγ , in
the vicinity of the resonance. Markers are from tracking (656 particles), solid curve and numerical
values of .Gγ R and .ν Z are from a fitting using Eq. 8.49

(i) Bunch depolarization.

Spin depolarizing resonances in SATURNE 1 are located at (Figs. 8.43 and 8.44)

. Gγ R = k ± ν Z = k ± 0.888284 ≡ 4 − 0.888284, 4 + 0.888284, 8 − 0.888284

where .ν Z has been taken from Table 8.8, or from Fig. 8.37. .Gγ R is bounded by
Gγ (3 GeV) = 7.525238 < 8 + ν Z
.
The simulation data file to track through these resonances is the same as in ques-
tion (e), Table 8.12, except for the following:
– substitute MCOBJET (to be uncommented) to OBJET (to be commented),
– under CAVITE substitute a peak voltage .V = 20 kV to .V = 200 kV,
– under REBELOTE, request a 300,000 turn cycle rather than 30,000.
MCOBJET creates a 200 particle bunch with Gaussian transverse and longitudinal
densities, with the following rms values at 3.6 MeV:

dp
ε /π = 25 µm, ε Z /π = 10 µm,
. Y = 10−4
p

CAVITE accelerates that bunch from 3.6 MeV to 3 GeV at a rate of .q V̂ sin(φs ) =
10 keV/turn (.V̂ = 20 kV , .φs = 150◦ ), in 300,000 turns.
Figure 8.43 shows sample . S Z spin components of a few particles taken among
the 200 tracked. Figure 8.44 displays .(S Z ), the average polarization of the bunch
(gnuplot script given in Table 8.16).
The strength of any one of the three resonances crossed can be computed, from
the upstream and downstream bunch polarization averaged over the 200 particles,
298 8 Weak Focusing Synchrotron

Fig. 8.43 Vertical spin component of a few particles accelerated from 3.6 MeV to 3 GeV. A graph
obtained using zpop: menu 7; 1/2 to open b_zgoubi.fai; 2/[39,23] for . S Z versus turn; 7 to plot

Fig. 8.44 Average vertical spin component of a 200 particle bunch, accelerated from 3.6 MeV to
3 GeV

using Eq. 8.58. Dependence upon the vertical emittance of the bunch can be per-
formed repeating this tracking simulation, with a different vertical emittance (under
MCOBJET).
Checking dependence upon crossing speed of the depolarizing effect of the res-
onances can be performed by repeating this tracking simulation with a different
accelerating rate .V̂ sin(φs ).
8.4 Solutions of Exercises of This Chapter: Weak Focusing Synchrotron 299

Table 8.16 A gnuplot script to plot the average vertical spin component of the 200 particle set,
along the acceleration ramp (Fig. 8.44). The average is prior computed by an awk script, which
reads the necessary data from zgoubi.fai.

8.2 Construct the ZGS Synchrotron. Spin Resonances

(a) A model of ZGS synchrotron.

DIPOLE is used to simulate both cell dipoles. It is necessary to have Fig. 8.28 at
hand (in addition to the users’ guide), when filling up the data list under DIPOLE.
Some comments regarding these data:

• DIPOLE field is defined in a cylindrical coordinate system.

• The bending sector is 45.◦ , this is also the field region extent angle . AT in the
preliminary hard-edge model.
• When accounting for fringe fields, the angular extent . AT has to encompass the
fringe fields, at both ends of the 45.◦ sector: an extra 5.◦ takes care of that, for a
total . AT =55.◦ , which ensures absence of truncation of the fringe fields at the . AT
sector boundaries, over the all radial excursion of the beam.
• . R M is given the curvature radius value, . R M = Bρ/B = 1.035270[T m] /
0.04986851[T] = 20.76 m, this makes magnet positioning and closed orbit checks
easier (see below).
• The field and reference rigidity are for injection energy, 50 MeV, an arbitrary
choice.
• ACENT=27.5.◦ is the reference azimuth for the positioning of the entrance and
exit EFBs. It is taken in the middle of the . AT range, an arbitrary choice.
300 8 Weak Focusing Synchrotron

• The entrance radius in the . AT sector is . R E = R M/ cos(AT − ω+ ) = R M/

cos(5◦ ), with .ω+ = 22.5◦ the positioning of the entrance EFB with respect to
ACENT (Fig. 8.28). And similarly for the positioning of the exit reference frame,
− ◦ − ◦
. RS = R M/ cos(AT − (AC E N T − ω )) = R M/ cos(5 ) with.ω = −22.5 the
+ − ◦
positioning of the exit EFB. Note that.ω − ω = 45 , the value of the bend angle.
• The entrance angle TE identifies with the extension to the 45.◦ sector, namely,
TE=-5.◦ . And similarly for the positioning of exit frame, 5.◦ downstream of the exit
EFB, TS=5.◦ .

In order to build the cell, and in the first place the two cell dipoles (they are mirror
symmetric, thus build one, the other follows), it is a good idea to proceed by steps:
(i) first build a 45.◦ sector in the hard edge model (Table 8.17). Outcomes of
FAISCEAU located next to DIPOLE indicate that a trajectory entering DIPOLE at
radius . R = R M, normal to the EFB (thus, .Y0 = 0 and .T0 = 0 in OBJET), does exit
with Y .= 0 and T .= 0. Data validation at this stage can be performed by comparing
DIPOLE’s transport matrix computed with MATRIX, and the theoretical expectation
(after Eq. 14.6):
⎛ ⎞ ⎛ ⎞
cos α ρ sin α 0 0 0 ρ(1 − cos α) 0.7071 14.6795 0 0 0 6.0804
⎜ − ρ1 sin α cos α 0 0 0 sin α ⎟ ⎜ −0.03406 0.7071 0 0 0 0.7071 ⎟
⎜ ⎟ α=π/4, ⎜ ⎟
⎜ 0 0 1 ρα 0 0 ⎟ ρ=20.76 ⎜ 0 0 1 16.3048 0 0 ⎟
.T = ⎜
⎜
⎟
⎟ = ⎜
⎜
⎟
⎜ 0 0 0 1 0 0 ⎟ ⎜ 0 0 0 1 0 0 ⎟ ⎟
⎝ sin α 0 0 0 1 ρ(α − sin α) ⎠ ⎝ 0.7071 0 0 0 1 1.6253 ⎠
0 0 0 0 0 1 0 0 0 0 0 1

Table 8.17 Simulation input data file: a 45.◦ sector bend in the hard edge model. The reference
trajectory has equal entrance and exit positions, and opposite sign angles. It coincides with the
arc . R = R M. MATRIX computes the transport matrix of the dipole (bottom of this Table), for
comparison with the fringe field model
8.4 Solutions of Exercises of This Chapter: Weak Focusing Synchrotron 301

Table 8.18 Outcomes of the simulation file of Table 8.17

MATRIX computation outcomes from raytracing can be found for comparison in

Table 8.18.
(ii) next, add fringe fields, including the 5.◦ extensions that add to. AT (Table 8.19).
Negative drifts with length. R M tan(5◦ ) = 181.62646548 cm have been added at both
ends, so to recover the actual 45.◦ sector opening. A FIT procedure finds the field
value necessary for recovering the exact orbit deviation, as the latter is perturbed
when introducing fringe fields. Again, FAISCEAU allows checking the correctness
of DIPOLE data: exit coordinates come out to be Y .= 0 and T .= 0; however the path
across the dipole is changed under the effect of the fringe fields, thus its length: s .=
302 8 Weak Focusing Synchrotron

Table 8.19 Simulation input data file: ZGS 45.◦ sector bend, with entrance and exit EFBs wedge
angles and fringe fields. The reference trajectory has equal entrance and exit position, and opposite
sign angles. It runs closely to the arc . R = R M, not strictly coinciding with the latter due to the
fringe fields. MATRIX computes the transport matrix of the dipole, for comparison with the hard
edge model. Negative drifts with length . R M tan(5◦ ) = 181.62646548 cm are added to recover the
hard edge path length

1630.459 cm is slightly different, compared to the hard edge case (an arc of radius
radius . R M = 2076 cm and length 1630.487 cm)
(iii) next, add the EFB angles: the sector is closing (wedge angles .ε1 > 0 and
.ε2 > 0 by convention) thus the EFB tilt angle .θ under DIPOLE is positive at entrance,
8.4 Solutions of Exercises of This Chapter: Weak Focusing Synchrotron 303

negative at exit (Fig. 8.28). In order to reach proper wave number values (this is
addressed below), the wedge angles are taken to be .ε1 = 13◦ and .ε2 = 8◦ . These
considerations result in the following:
– the entrance (respectively exit) EFB of the upstream dipole of the cell is tilted with
respect to the reference orbit by an angle .θ = +13◦ (resp. .θ = −8◦ ) (Fig. 8.27),
– the entrance (resp. exit) EFB of the downstream dipole is tilted with respect to the
reference orbit by an angle .θ = +8◦ (resp. .θ = −13◦ ).
This final step requires again re-adjusting the radial positioning of the dipole (RE and
RS, entrance and exit radius respectively), and field. In that aim the FIT procedure
in Table 8.19 is added a variable: the RE and RS radii, coupled, and a constraint:
the reference orbit has zero radial excursion at exit of the dipole. This FIT results in
re-adjusted magnetic field and RE, RS positioning, with the respective values

. B0 = 0.49860858 kG and R E = RS = 2084.5090 cm

These are the values used in the ZGS cell simulation in Table 8.20,
(iv) and, finally, assemble this dipole and its mirror symmetric, in a cell (Fig. 8.27
and Table 8.20). The mirror symmetric is obtained by just permuting the entrance
and exit wedge angles. The cell includes a half long-drift at each end, and a short
drift between the dipoles. The three have been taken equal for simplification, 3.37 m
long.
Lattice parameters
The TWISS command down the sequence (Table 8.20) produces the periodic beam
matrix results shown in Table 8.21. It also produces a zgoubi.TWISS.out file which
details the optical functions along the sequence (at the downstream end of the optical
elements). The header of that file details the optical parameters of the structure
(Table 8.22).

(b) Betatron functions of the ZGS cell.

Among the various ways to produce the betatron functions along the sequence
(and throughout the DIPOLEs), here are two possibilities, based on the storage of
particle coordinates in zgoubi.plt during stepwise raytracing:
1. a direct way consists in using OBJET[KOBJ=5] and transport the 13-particle set
so obtained across the sequence; then, betaFromPlt from zgoubi toolbox [39] can
be used to compute the transport matrix from the origin, step by step along the
sequence, from particle coordinate values logged in zgoubi.plt during the stepwise
integration;
2. an indirect way consists in launching a few particles on a common invariant (hori-
zontal and/or vertical) and subsequently plot the s-dependent quantities .Ŷ 2 (s)/εY
and/or . Ẑ 2 (s)/ε Z . The maximum value of the latter, a function of the distance s,
is the betatron function along the sequence, .βY,Z (s).
The second method is used here (this is an arbitrary choice. Exercises may be
found in the various Chapters, that use the first method and may be referred to).
304 8 Weak Focusing Synchrotron

Table 8.20 Simulation input data file: ZGS cell simplified model, obtained by assembling DIPOLE
taken from Table 8.19 and its mirror symmetric, and adding drift spaces. This input data file defines
the ZGS cell sequence segment S_ZGS_cell to E_ZGS_cell. It also defines the dipole segments
S_ZGS-DIP_UP to E_ZGS-DIP_UP (first dipole of the cell) and S_ZGS-DIP_DW to E_ZGS-
DIP_DW (second dipole of the cell). In further INCLUDE statements, this file is used under the
name ZGS_cell.inc
8.4 Solutions of Exercises of This Chapter: Weak Focusing Synchrotron 305

Table 8.21 Results obtained running the simulation input data file of Table 8.20, ZGS cell—an
excerpt from zgoubi.res execution listing

Table 8.22 An excerpt of zgoubi.TWISS.out file resulting from ZGS cell simulation of Table 8.20.
Note that the ring (4-period) wave numbers are 4 times the cell values Q1, Q2 below. Optical
functions (betatron function and derivative, orbit, phase advance, etc.) along the optical sequence
are listed as part of zgoubi.TWISS.out following the header. The top and bottom parts of that listing
are given below
306 8 Weak Focusing Synchrotron

Table 8.23 Simulation input data file: raytrace 60 particles across ZGS cell to generate beam
envelopes. Store particle data in zgoubi.plt, along DIPOLEs and split DRIFTs. The INCLUDE file
and segments are defined in Table 8.20

The input data file to derive the betatron function following method (2) above
is given in Table 8.23. The initial ellipse parameters (under OBJET) are the peri-
odic values, namely, .αY = α Z = 0, .βY = 28.63 m, .β Z = 37.01 m, they are a sub-
product of the TWISS procedure performed in (a), to be found in zgoubi.TWISS.out
(Table 8.22). The resulting envelopes and their squared value are shown in Fig. 8.45.
Note that this raytracing also provides the coordinates of the 60 particles on their
common upright invariant
,2
. x /β x + β x x = εx /π
2

at start and at the end of the cell (with x standing for either .Y or . Z , and .εY,Z /π =
10−4 , here). This allows checking that the initial ellipse parameters (under OBJET,
Table 8.23) are effectively periodic values, and that the raytracing went correctly,
namely by observing that the initial and final ellipses do superimpose (Fig. 8.46).
Dispersion function
Raytracing off-momentum particles on their chromatic closed orbit provides the
periodic dispersion function. In order to do so, the input data file of Table 8.23 can
be used, it just requires changing OBJET to the following:
8.4 Solutions of Exercises of This Chapter: Weak Focusing Synchrotron 307

Fig. 8.45 Left: horizontal and vertical envelopes as generated by plotting the coordinates Y(s) (thick
lines, red, along the drifts only) or Z(s) (thin lines, blue) across the ZGS cell, of 60 particles evenly
distributed on a common .10−4 π m invariant, either horizontal or vertical (while the other invariant
is zero). Right: a plot of .Y 2 (s)/εY and . Z 2 (s)/ε Z : the extrema identify with .βY (s) and .β Z (s),
respectively. The extrema extremorum values are .β̂Y = 32.5 m and .β̂ Z = 37.1 m, respectively.
Graphs obtained using zpop: menu 7; 1/1 to open zgoubi.plt; 2/[6,2] (or [6,4]) for .Y versus .s (or
. Z versus .s); 7 to plot; option 3/14 to raise Y (or Z) to the square and normalize to .εY,Z /π

Fig. 8.46 Sixty particles evenly distributed on a common periodic invariant (of value either .εY =
10−4 π m and .ε Z = 0, left graph, or the reverse, right graph) have been tracked from start to end of
the cell. These periodic invariants are defined assuming the periodic ellipse parameters determined
from prior TWISS, given in Table 8.22; values resulting from an rms match of the coordinates are
given in the figure, and do agree with those TWISS data. The figure shows the good superposition
of the start and end invariants (the start and end rms match ellipse parameters show negligible
difference), which confirms the correct value of the periodic ellipse parameters, namely, left graph:
horizontal phase space at start (crosses) and end (dots) of the cell; right graph: vertical phase space
at start (crosses) and end (dots) of the cell
308 8 Weak Focusing Synchrotron

Fig. 8.47 A graph of the radial excursion, within DIPOLE range (namely, . AT = 55◦ extent,
Table 8.20), of an on-momentum particle (its radial position in the dipole body is . R0 ≈ 20.7628 m,
corresponding to Y=0 in this graph) and two particles at respectively.dp/ p = ±10−3 . The diverging
parts at DIPOLE ends are in the 5.◦ fringe field regions. A graph obtained using zpop: menu 7; 1/1
to open zgoubi.plt; 2/[6,2] for .Y versus distance; 7 to plot

The position and angle of the chromatic particles, which are offset by ./\p/ p =
±10−3 , are drawn from the value of the periodic dispersion .ηY = 36.85 m and
its derivative .ηY, ≈ 0 (Table 8.22), namely, .Y0 = ηY /\p/ p = ±3.685 cm and .T0 =
ηY /\p/ p = 0.
Running Table 8.23 simulation with this new OBJET produces the following
coordinates at FAISCEAU, located at the end of the sequence (an excerpt from
zgoubi.res execution listing):

The local coordinates Y, T (under FAISCEAU, right hand side) are equal to the
initial coordinates .Y0 , .T0 (under OBJET, left hand side), to better than .5 µm, .0.5 µrad
accuracy respectively (zgoubi.fai can be consulted for double precision on these
values), so confirming the periodicity of these chromatic trajectories. Figure 8.47
shows the particle trajectories through the two DIPOLEs. A difference between the
on- and off-momentum trajectories yields as expected a quasi-constant .ηY ≈ 36.8 m
whereas .ηY, ≈ 0.
8.4 Solutions of Exercises of This Chapter: Weak Focusing Synchrotron 309

Fig. 8.48 Dispersion

function along ZGS cell,
obtained by orbit difference.
The discontinuities are
artifacts, they are located in
the overlapping regions
between the optical sequence
DIPOLEs and DRIFTs
(Table 8.23)

Orbit difference
The method can be used to compute the dispersion function. This requires tracking
particles with .±dp/ p momentum offset. A gnuplot script can compute and plot the
orbit difference, and normalize to dp/p; the result is the periodic dispersion, displayed
in Fig. 8.48.

(c) Some verifications regarding the model.

The field along large excursion orbits can be logged in zgoubi.plt, using option
IL=2 (or 20, or 200, etc. for printout every 10, or 100, etc. integration step) under
DIPOLE.
The simulation file of Table 8.23 is used to raytrace five particles, with OBJET
changed to the following:

Apart from the on-momentum particle (2nd in the list) this OBJET defines two
particles on./\p/ p = ±1% chromatic orbit (1st and 3rd in the list), this is an excursion
of a few tens of centimeters, large as requested, as ./\x ≈ 38 × dp/ p. OBJET also
defines 2 particles launched into the cell at respectively . Z 0 = 5 cm and . Z 0 = 20 cm.
The magnetic field as a function of the azimuthal angle in DIPOLE frame, along
these trajectories across the upstream DIPOLE of the cell, is shown in Fig. 8.49. The
field curves for the first four trajectories essentially superimpose except for the fringe
field regions (Fig. 8.49), due to the wedge angles. This behaves as expected. Detailed
310 8 Weak Focusing Synchrotron

Fig. 8.49 Magnetic field along 5 different trajectories across the upstream DIPOLE, including four
large horizontal and vertical excursion cases, and a zoom in on the entrance fringe field region

inspection is possible, from particle coordinate and field data in zgoubi.plt—this is

out of the scope of the present question.
The field along the 5th particle trajectory features overshoots (Fig. 8.49), this is
due to the large vertical excursion (. Z ≈ 20 cm in the entrance fringe field region).
It looks reasonable, however it may be an artifact in the case that the high order
derivatives of the field in that region are large, resulting from the truncated Taylor
series method used for off mid-plane field extrapolation [36, Sect. 1.3.3].

(d) Sinusoidal approximation of the betatron motion.

The approximation
. y(θ ) = A cos(ν Z θ + φ)

is checked here considering the vertical motion (considering the horizontal motion
leads to similar conclusions). The value of the various parameters in that expression
are determined as follows:
– the particle raytraced for comparison is launched with an initial excursion
. Z 0 (θ = 0) = 5 cm (4th particle in OBJET, above). At the launch point (middle of
the long drift) the beam ellipse is upright (see below), whereas phase space motion
is clockwise, thus take
.A = 5 cm and φ = π/2

– the vertical betatron tune of the 4-cell ring is (Table 8.22)

ν = 4 × 0.192869 = 0.77147
. Z
8.4 Solutions of Exercises of This Chapter: Weak Focusing Synchrotron 311

Fig. 8.50 Vertical betatron motion, five turns around the ZGS ring, from raytracing (continuous
line), and sine approximation, superimposed (dashed line)

{
– .θ = s/R and . R = ds/2π with (Table 8.22)

.2π R = circumference = 4 × 42.72614331 = 170.90457 m

Consistency with sinusoidal approximation is shown in Fig. 8.50.

(e) An acceleration cycle. Symplecticity checks.

Eleven particles are launched for 65,000 turn tracking at a rate of

./\W = q V̂ sin φs = 400 × sin 1500 = 200 keV/turn

(. E : 0.05 → 13.05 GeV), all evenly distributed on the same initial vertical invariant

. Z 2 /β Z + β Z Z ,2 = ε Z /π (8.60)

with .ε Z /π = 10−4 m, or, normalized, .βγ ε Z /π = 0.33078 × 10−4 m.

The simulation file is given in Table 8.24. CAVITE[IOPT=3] is used, it provides
an RF phase independent boost

./\W = q V̂ sin φs

as including synchrotron motion is not necessary here, even better, anticipating on

spin question (f) this ensures constant depolarizing resonance crossing speed, so
precluding any possibility of multiple crossing (it can be referred to [40] regarding
that effect).
312 8 Weak Focusing Synchrotron

Table 8.24 Simulation input data file: track 11 particles launched on the same vertical invariant,
with zero horizontal invariant. The INCLUDE adds the ZGS cell four times, the latter is defined in
Table 8.20. An MCOBJET is commented, it is used in a subsequent spin tracking exercise

Betatron damping
Figure 8.51 shows the damped vertical motion of the individual particles, over the
acceleration range, together with the initial and final distributions of the 11 parti-
cles on elliptical invariants. Departure from the matching ellipse at the end of the
acceleration cycle, 13 GeV (Eq. 8.60 with .ε Z /π = 2.2244 × 10−7 m), is marginal.
Degree of non-symplecticity of the numerical integration
The degree of non-symplecticity as a function of integration step size is illustrated
in Fig. 8.52. The initial motion is taken paraxial, vertical motion is considered as it
resorts to off-mid plane Taylor expansion of fields [36, DIPOLE Sect.], a stringent
test as the latter is expected to deteriorate further the non-symplecticity inherent
to the Lorentz equation integration method (a truncated Taylor series method [36,
Eq. 1.2.4]).
8.4 Solutions of Exercises of This Chapter: Weak Focusing Synchrotron 313

Fig. 8.51 Left: damped vertical motion, from 50 MeV to 13.05 GeV, 65,001 turns. Right: the ini-
tial coordinates of the 11 particles (squares) are taken on a common invariant .ε Z (0) = 10−5 π m
(at 50 MeV, .βγ = 0.33078, thus .βγ ε Z (0) = 0.33078 × 10−5 π m ); the final coordinates after
65,000 turns (crosses) appear to still be (with negligible departure) on a common invariant
of value .ε Z ( f inal) = 2.2244 × 10−7 π m, or .βγ ε Z ( f inal) = 0.33076 × 10−5 π m (at 13 GeV,
.βγ = 14.869842), equal to the initial value

Fig. 8.52 Turn-by-turn evolution of the normalized invariant,.βγ ε Z (tur n)/βγ ε Z (0) (initial.ε Z (0)
taken paraxial), for integration step sizes 1, 2 and 4 cm

Evolution of the wave numbers

The Fortran tool tunesFromFai_iterate can be used to computes tunes as a function
of turn number or energy, it reads turn-by-turn particle data from zgoubi.fai and
computes a discrete Fourier transform over so many turns (a few tens, for instance),
every so many turns [41]. Typical results are displayed in Fig. 8.53, tunes have the
expected values: .νY = 0.849, .ν Z = 0.771. An acceleration rate of 200 keV/turn has
been taken (namely, .V̂ = 400 kV and still .φs = 1500 ), to save on computing time.
Note that turn-by-turn raytracing allows determining the tune value at all .γ along
the acceleration cycle (and thus for instance the .γ values at which the resonance
314 8 Weak Focusing Synchrotron

Fig. 8.53 Horizontal ring tune (left vertical axis),.νY ≈ 0.8494, and vertical ring tune (right vertical
axis), .ν Z ≈ 0.77147, as a function of turn number, over 65,000 turns (. E : 0.05 → 13 GeV at a rate
of 200 keV/turn). The graph displays results for integration step sizes 1, 2 and 4 cm, essentially
converged

occurs, see (f)). In these simulations anyway the horizontal and vertical tunes are
essentially constant over the all cycle: it is determined by the wedge angle, which
will not charge as long as the reference orbit is not changed. The latter holds here,
as SCALING[NTIM=-1] causes the magnet field to strictly follow the momentum
boost by CAVITE.

(f) Crossing an isolated intrinsic depolarizing resonance.

The simulation uses the input data file of Table 8.24, with the following changes:

• Under OBJET:

– 1st line, change the reference rigidity. B O R O to the proper value, a few thousand
turns upstream of the resonance to be crossed;
– 3rd line, request a single particle (“1 1 1”, in lieu of “1 11 1”);
– 6th line, set the invariant .ε Z /π to the desired value, .εY /π value is indifferent;

• change the field value under DIPOLE consistently with the new BORO value, so
to maintain the expected curvature radius .ρ0 = B O R O/B = 20.76 m (Table 8.2);
• under CAVITE, provide the desired peak voltage .V̂ ;
• under REBELOTE, set the number of turns: a few thousands of turns upstream
and downstream of the resonance.

Similar simulations are performed in questions (f)–(i) of Exercise 8.1. Please refer
to the solutions of these SATURNE 1 simulations.
8.4 Solutions of Exercises of This Chapter: Weak Focusing Synchrotron 315

(g) Study of an imperfection depolarizing resonance.

The simulation data files of question (f) can be used here, mutatis mutandis, and
the methodology in (f) can be followed.
Similar simulations are proposed in questions (f)–(i) of Exercise 8.1, as well as in
the “Strong Focusing Synchrotron” Chapter exercises. Please refer to the solutions
of these exercises.

(h) Spin tracking. Bunch polarization.

Spin depolarizing resonances in the ZGS are located at

. Gγ R = k P ± ν Z = 4 − ν Z , 4 + ν Z , 8 − ν Z , 8 + ν Z , 12 − ν Z , etc.

with P .= 4 the superperiodicity of the ring, and .ν Z = 0.77147 taken from Table 8.22,
or from Fig. 8.53. .Gγ R is bounded, in the present simulation, by .Gγ (17.4 GeV ) =
35.0 < 9P − ν Z . Resonances are expected to be stronger at .Gγ R = 2 × 4k ± ν Z =
8 − ν Z , 8 + ν Z , 16 − ν Z , etc., with the additional factor 2 the number of cells per
superperiod [32, Sect. 3.II].
The simulation data file to track through these resonances is the same as in ques-
tion (e), Table 8.24, except for the substitution of MCOBJET (to be uncommented) to
OBJET (to be commented). MCOBJET creates a 200 particle bunch with Gaussian
transverse and longitudinal densities, with the following rms values at 50 MeV:

dp
ε /π = 25 µm, ε Z /π = 10 µm,
. Y = 10−4
p

which are presumably close to ZGS polarized proton runs [33]. CAVITE accelerates
that bunch from 50 MeV to 17.4 GeV about, at a rate of .q V̂ sin(φs ) = 200 keV/turn
(.V̂ = 400 kV , .φs = 30◦ ), in 87,000 turns about.
Figure 8.54 shows sample . S Z spin components of a few particles taken among
the 200 tracked. Figure 8.55 displays .(S Z ), the vertical polarization component of
the bunch (gnuplot script given in Table 8.16).

8.3 Visible SR from GEC 70 MeV Synchrotron

The input data file to simulate an electron trajectory in GEC dipole is given in
Table 8.25.
The critical energy is .ωc /2π = 3γ 3 c/(4πρ) = 2.6635 eV, a critical wavelength
of .λc = 0.4655 µm, in the visible range. The GEC glass vacuum chamber was trans-
parent, this allowed the observation of this synchrotron light.
The rms angular aperture of the radiation is .∼ 1/γ = 7.247 rad. By taking a 22 cm
long magnet, the cone captured by observing tangentially to the arc extends over
./\φ = 22[cm] /(ρ/γ ) ≈ ±52 times the SR angular aperture .1/γ , well beyond the
peak region of the electric impulse. Thus this truncature only concerns quasi-zero
tail regions of the impulse (Fig. 8.17), its Fourier transform can be considered accurate
enough.
316 8 Weak Focusing Synchrotron

Fig. 8.54 Individual vertical spin component of 20 particles accelerated in ZGS from 50 MeV to
17.4 GeV, at a rate of 200 keV/turn. A graph obtained using zpop: menu 7; 1/2 to open b_zgoubi.fai;
2/[20,23] for . S Z versus energy; 7 to plot

Fig. 8.55 Average vertical component of the polarization vector of a 200 particle bunch, acceler-
ated from 50 MeV to 17.4 GeV. The vertical lines materialize the locations .Gγ R = 4k ± ν Z of the
depolarizing resonances. Resonances are strongest at .Gγ R = 8k ± ν Z (as labeled)
References 317

Table 8.25 Simulation input data file: computation of the 29.2 cm radius trajectory of a 70 MeV
electron in a dipole field

The two components of the electric field impulse in the direction .(φ, ψ) =
(0, 0.1/γ ), and their spectral brightness, are displayed in Fig. 8.17. The .σ com-
ponent peaks near .hωc = 2.6635 eV, as expected.

References

1. V.I. Veksler, A new method of accelerating relativistic particles. Comptes-Rendus de

l’Académie des Sciences de l’URSS 43(8), 329–331 (1944)
2. E.M. McMillan, The synchrotron. Phys. Rev. 68, 143–144 (1945)
3. F.K. Goward, D.E. Barnes, Experimental 8 MeV synchrotron for electron acceleration. Nature
158, 413 (1946)
4. H.C. Pollock, The discovery of synchrotron radiation. Am. J. Phys. (1983)
5. F.R. Elder et al., Radiation from electrons in a synchrotron. Phys. Rev. 71, 829–Published 1
June 1947
6. D.W. Kerst, The acceleration of electrons by magnetic induction. Phys. Rev. 60, 47–53 (1941)
7. SATURNE 1 photos: credit CEA Saclay. Archives historiques du CEA. Copyright CEA/Service
de documentation
8. A. Sessler, E. Wilson, Engines of Discovery. A Century of Particle Accelerators (World Sci-
entific, 2007)
9. Fig. 8.3 : Credit Reider Hahn, Fermilab
10. K. Endo et al., Compact proton and carbon ion synchrotrons for radiation therapy. MOPRI087,
in Proceedings of EPAC 2002, Paris, France, pp. 2733–2735. https://accelconf.web.cern.ch/
e02/PAPERS/MOPRI087.pdf
11. V.A. Vostrikov, et al., Novel approach to design of the compact proton synchrotron magnetic
lattice. tupsa17, 26th Russian Particle Accelerator Conference RUPAC2018, Protvino, Russia
(2018). https://accelconf.web.cern.ch/rupac2018/papers/tupsa17.pdf
12. K.R. Symon, MURA Days, in Proceedings of the 2003 Particle Accelerator Conference. https://
accelconf.web.cern.ch/p03/PAPERS/WOPA003.PDF
13. T. Ohkawa, Two-beam fixed field alternating gradient accelerator. Rev. sci. Instrum. 29, 108–17,
1
14. Bernardini, C., AdA: the first electron-positron collider. Phys. Perspect. 6 (2004). 156–183
1422-6944/04/020156-28. https://link.springer.com/article/10.1007/s00016-003-0202-y
15. Image by Argonne National Laboratory, Comm (L.A, Martinez, ANL, Apr. 2023)
318 8 Weak Focusing Synchrotron

16. D. Cohen, Feasibility of accelerating polarized protons with the argonne ZGS. Rev. Sci. Instrum.
33, 161 (1962). https://doi.org/10.1063/1.1746524
17. L.G. Ratner, T.K. Khoe, Acceleration of polarized protons in the zero gradient synchrotron,
in Procs. PAC 1973 Conference, Washington (1973). http://accelconf.web.cern.ch/p73/PDF/
PAC1973_0217.PDF
18. J. Bywatwr, T. Khoe et al., A pulsed quadrupole system for preventing depolarization. IEEE
Trans. Nucl. Sci. 20(3) (1973)
19. Y. Cho et als., Effects of depolarizing resonances on a circulating beam of polarized protons
during or storage in a synchrotron. IEEE Trans. Nucl. Sci. NS-24(3) (1977)
20. E.F. Parker, High energy polarized deuterons at the argonne national laboratory zero gradient
synchrotron. IEEE Trans. Nucl. Sci. NS-26(3), 3200–3202 (1979)
21. D.E. Suddeth et als., Pole face winding equipment for eddy current correction at the zero
gradient synchrotron, in Procs. PAC 1973 Conference, Washington (1973). http://accelconf.
web.cern.ch/p73/PDF/PAC1973_0397.PDF
22. A.V. Rauchas, A.J. Wright, Betatron tune profile control in the zero gradient synchrotron (ZGS)
using the main magnet pole face windings, in Procs. PAC1977 Conference, IEEE Transactions
on Nuclear Science, NS-24(3) (1977)
23. G. Floquet, Sur les équations différentielles linéaires à coefficients périodiques. Annales sci-
entifiques de l’E.N.S. 2e série, tome 12, pp. 47–88 (1883). http://www.numdam.org/item?
id=ASENS_1883_2_12__47_0
24. G. Leleux, Accélérateurs Circulaires Lecture Notes (INSTN, CEA Saclay, 1978)
25. T. Risselada, Transition gamma jump schemes, in Proceedings of Jyv askyl a CERN Accelerator
School, 7–18 Sept. 1992. Yellow Report CERN 94-01
26. A. Hofmann, The physics of synchrotron radiation. Cambridge monographs on particle physics,
in Nuclear Physics and Cosmology (20) (Cambridge University Press, 2004)
27. F. Méot, A theory of low frequency far-field synchrotron radiation. Part. Accel. 62, 215–239
(1999)
28. F. Méot, L. Ponce, N. Ponthieu, Low frequency interference between short SR sources. PRST-
AB 4, 062801 (2001)
29. F. Méot, Zgoubi Users’ Guide. https://www.osti.gov/biblio/1062013-zgoubi-users-guide.
Sourceforge latest version. https://sourceforge.net/p/zgoubi/code/HEAD/tree/trunk/guide/
Zgoubi.pdf
30. G. Leleux, Traversée des résonances de dépolarisation. Rapport Interne LNS/GT-91-15, SAT-
URNE, Groupe Théorie, CEA Saclay (février, 1991)
31. F. Méot, Spin dynamics. In: Polarized beam dynamics and instrumentation in particle accel-
erators, in USPAS Summer 2021 Spin Class Lectures (Springer Nature, Open Acess, 2023).
https://link.springer.com/book/10.1007/978-3-031-16715-7
32. S.Y. Lee, Spin Dynamics and Snakes in Synchrotrons (World Scientific, 1997)
33. T.K. Khoe et al., The high energy polarized beam at the ZGS, in Procs. IXth International
Conference on High Energy Accelerators, Dubna, pp. 288–294 (1974). Figure 8.19: Copy-
rights under license CC-BY-3.0. https://creativecommons.org/licenses/by/3.0; no change to
the material
34. M. Froissart, R. Stora, Dépolarisation d’un faisceau de protons polarisés dans un synchrotron.
Nucl. lnst. Meth. 7, 297 (l960)
35. H. Bruck, P. Debraine, R. Levy-Mandel, J. Lutz, I. Podliasky, F. Prevot, J. Taieb, S.D. Winter,
R. Maillet, Caractéristiques principales du Synchrotron à Protons de Saclay et résultats obtenus
lors de la mise en route, rapport CEA no. 93, CEN-Saclay (1958)
36. F. Méot, Zgoubi Users’ Guide. https://www.osti.gov/biblio/1062013-zgoubi-users-guide
Sourceforge latest version. https://sourceforge.net/p/zgoubi/code/HEAD/tree/trunk/guide/
Zgoubi.pdf
37. M.H. Foss et al., The argonne ZGS magnet. IEEE 1965, 377–382 (1965)
38. L.A. Klaisner et al., IEEE 1965, 133–137 (1965)
39. A post-processing tool to transport betatron functions step-by-step, using raytracing data stored
in zgoubi.plt. https://sourceforge.net/p/zgoubi/code/HEAD/tree/trunk/toolbox/betaFromPlt/
References 319

40. T. Aniel et al., Polarized particles at SATURNE. Journal de Physique, Colloque C2, supplément
au n02, Tome 46, février, pp. C2–499 (1985). https://hal.archives-ouvertes.fr/jpa-00224582
41. The Fortran tunesFrmFai_iterate.f, together with a README and an example of its use, can
be found at https://sourceforge.net/p/zgoubi/code/HEAD/tree/trunk/toolbox/tunesFromFai/
42. https://stackoverflow.com/questions/42677017/plot-average-of-nth-rows-in-gnuplot
43. F. Méot, Spinor methods, in Polarized Beam Dynamics and Instrumentation in Particle Accel-
erators. USPAS Summer 2021 Spin Class Lectures (Springer Nature, Open Access, 2023).
https://link.springer.com/book/10.1007/978-3-031-16715-7

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0
International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing,
adaptation, distribution and reproduction in any medium or format, as long as you give appropriate
credit to the original author(s) and the source, provide a link to the Creative Commons license and
indicate if changes were made.
The images or other third party material in this chapter are included in the chapter’s Creative
Commons license, unless indicated otherwise in a credit line to the material. If material is not
included in the chapter’s Creative Commons license and your intended use is not permitted by
statutory regulation or exceeds the permitted use, you will need to obtain permission directly from
the copyright holder.
Chapter 9
Strong Focusing Synchrotron

Abstract This chapter introduces the strong focusing alternating gradient (AG)
and separated function synchrotrons. It provides the theoretical material which the
simulation exercises lean on. The chapter begins with a brief reminder of the his-
torical context, and continues with beam optics, chromaticity, acceleration, reso-
nances and resonant extraction, dynamical effects of synchrotron radiation (SR),
the electromagnetic SR impulse, and depolarizing resonances. This resorts to basic
charged particle optics, acceleration, and dynamics in magnetic fields introduced in
the previous chapters. The simulation of a strong focusing AG synchrotron requires
just two optical elements from zgoubi library: DIPOLE or MULTIPOL to sim-
ulate a combined function dipole, and DRIFT to simulate straight sections. Main
dipoles in a separated function synchrotron can use BEND. It requires in addition
quadrupoles, simulated using QUADRUPO or MULTIPOL. The latter can simulate
higher order lenses, which can otherwise resort to SEXTUPOL, OCTUPOLE, etc.
Acceleration uses CAVITE. Accounting for synchrotron radiation (SR) energy loss
requires SRLOSS. Monte Carlo SR monitoring can use SRPRNT, which logs data in
zgoubi.res. SRPRNT[PRINT] in addition logs data in zgoubi.SRPRNT.Out. Com-
putation of synchrotron radiation (SR) Poynting and spectral brightness uses zpop.
Particle monitoring requires keywords introduced in the previous Chapters, including
FAISCEAU, FAISTORE, possibly PICKUPS, and some others. Spin motion com-
putation and monitoring resort to SPNTRK, SPNPRT, FAISTORE. Optics matching
and optimization use FIT[2]. INCLUDE is used, mostly here in order to simplify the
input data files. SYSTEM is used to, mostly, resort to gnuplot so as to end simula-
tions with some specific graphs. Data for the latter are read from output files filled
up during the execution of the code, such as zgoubi.fai (resulting from the use of
FAISTORE), zgoubi.plt (resulting from IL.=2), or other zgoubi.*.out files resulting
from a PRINT command. Stepwise particle data logged in zgoubi.plt are used by the
interface zpop to compute the electric field impulse of SR and subsequent spectral
angular energy density of the radiation.

© The Author(s) 2024 321

F. Méot, Understanding the Physics of Particle Accelerators, Particle Acceleration
and Detection, https://doi.org/10.1007/978-3-031-59979-8_9
322 9 Strong Focusing Synchrotron

Notations used in the Text

B; . Bx,y,s ; . B
. Field vector; its components in the moving frame; its modulus
Bρ = p/q; Bρ0
. Particle rigidity; reference rigidity
[ straight
C; .C0
. Orbit length; .C = 2π R + ; reference, .C0 = C
sections
( p = p0 )
.E; . E σ , . E π SR electric field impulse; its parallel and normal components
. E; E s Particle energy, . E = γm 0 c2 ; synchronous energy
EFB Effective Field Boundary
. f rev , . f rf = h f rev Revolution and RF voltage frequencies
.G Gyromagnetic anomaly, .G = 1.792847 for proton
. G; . K = G/Bρ Quadrupole gradient; focusing strength
.h RF harmonic number
.m; .m 0 ; . M Particle mass; rest mass; in units of MeV/c.2
ρ ∂B
.n = − Focusing index
B ∂x
.n0 Stable spin precession direction
.P = E × B SR Poynting vector
. Pi , P f Beam polarization, initial, final
.p; . p; . p0 Momentum vector; its modulus; reference
.q Particle charge
.r ; R Orbital radius; average radius, . R = C/2π
.S Periodicity of the lattice
.s Path variable
.Us SR energy loss
.v; .v Particle velocity vector; its modulus
. V (t); V̂ Oscillating voltage; its peak value
, ,
[ ]
.x, x , y, y , l,
dp
p
Particle coordinates in the moving frame, . (∗), = d(∗)/ds
.α Momentum compaction; or trajectory deviation; or depolarizing
resonance crossing speed
.β = v/c; .β0 ; .βs Normalized particle velocity; reference; synchronous
β
. u Betatron functions (.u : x, y, Y, Z )
.γ = E/m 0 c
2
Lorentz relativistic factor
√
.γtr Transition .γ, .γtr = 1/ α
.δ p, ./\p Momentum offset
.ec Critical energy of SR, .ec = hωc = hc/λc
.ε Wedge angle
.εu /π Courant-Snyder invariant; emittance/./π (.u : x, y, l)
.e R Strength of a depolarizing resonance
.η Phase slip factor, .η = γ12 − α
{
.μu Betatron phase advance per period, .μu = period βuds(s) (.u : x, y)
.νu Wave numbers, horizontal, vertical, synchrotron (.u : x, y, l)
.ρ; ρ0 Curvature radius; reference
.σ Beam matrix
9.1 Introduction 323

.φ; φs Particle phase at voltage gap; synchronous

{ phase
.ϕu Betatron phase advance, .ϕu = ds/βu (.u : x, y, Y, or Z )
.ϕ Spin angle to the vertical axis
.ωc Critical angular frequency of SR, .ωc = 3γ 3 c/2ρ
.ωs ; .gs .2π f rev ; synchrotron frequency

9.1 Introduction

In the very manner that the 1930s–1940s cyclotron, betatron, microtron, weak focus-
ing synchrotron, which are still in use today, have since essentially not changed in
their concepts and design principles, today the gap profile, yoke and current coil
geometry of combined function alternating-gradient (AG) dipoles remain essentially
as patented in 1950 (Fig. 9.1) [1].
In 1952, in the context of studies concerning the Cosmotron, strong focusing
was devised at the Brookhaven National Laboratory (BNL): “Strong focusing forces
result from the alternation of large positive and negative .n -values in successive sectors
of the magnetic guide field in a synchrotron. This sequence of alternately converg-
ing and diverging magnetic lenses [...] leads to significant reductions in oscillation
amplitude ” [3]. It led to the construction of the first two high-energy AG proton
synchrotrons (PS), in the 30 GeV range, in the late 1950s: the CERN PS, and the
AGS at BNL (Fig. 9.2). Both remain major pieces, 60 years later, of the respective
injection chains of the two largest colliders in operation, the LHC and RHIC. Early
works at BNL provided theoretical formalism, still at work today, for the analysis of
beam dynamics in synchrotrons [4].
Separated function focusing, whereby beam guiding is ensured by uniform field
dipoles while focusing is ensured separately by quadrupoles (Fig. 9.3), followed from
the development of the latter (Fig. 9.4), a spin-off of the strong index technology [7].
The dramatic reduction of transverse beam size by strong focusing allows guid-
ing and focusing magnets with small aperture, from lowest energies: medical syn-

Fig. 9.1 Bending magnet pole profiles for a focusing system for ions and electrons [1]. Assuming
curvature center to the left, the left (respectively right) profile is defocusing (resp. focusing), the
middle profile has zero index
324 9 Strong Focusing Synchrotron

Fig. 9.2 Top: the AGS combined function main dipole. The hyperbolic profile poles are visible,
partly hidden by the field coils. Bottom: the 809 m circumference AGS synchrotron, comprised of
240 such dipoles [2]

chrotrons in the 100 MeV range for instance, to highest ones: hundreds of GeV to
multi-TeV range particle physics and nuclear physics colliders (Fig. 9.5). Beams in
all these machines are essentially confined in a sub-centimeter or sub-millimeter
scale transverse space. A synchrotron is a string of dipole and multipole magnets
through which runs a vacuum pipe of a few centimeters diameter (hadron rings) or
a few millimeters (electrons). The size of the ring is essentially determined by its
circumference, proportional to the magnetic rigidity. This revolutionized the race
to high energies, from the prior few GeV weak focusing synchrotrons and their
huge magnets, to todays 7 TeV, 27 km long LHC and with further plans for 100 TeV,
9.1 Introduction 325

Fig. 9.3 SATURNE 2 strong focusing 3 GeV synchrotron at Saclay [5], successor in the late 1970s
of SATURNE 1 weak focusing synchrotron (Fig. 8.1). It was the first strong focusing synchrotron
to accelerate polarized ion beams

Fig. 9.4 A quadrupole

magnet at LBL in 1957, used
for beam lines at the
184-inch cyclotron. An early
specimen here, obviously,
being a spin-off of the early
1950s concept of strong
focusing [6]
326 9 Strong Focusing Synchrotron

Fig. 9.5 In RHIC tunnel at

the Brookhaven National
Laboratory [2]. The two
rings of the 255 GeV
polarized proton beams and
heavy ion collider run
parallel over 3.8 km, and
intersect at two experiments,
STAR and SPHENIX

Fig. 9.6 The ion rapid

cycling medical synchrotron
(iRCMS) [9], an ion beam
RCS for the treatment of
cancer tumors

100 km circumference colliders [8]. Strong focusing fostered the development of

high energy synchrotron light sources around the world, with high brightness syn-
chrotron radiation (SR) from UV to gamma rays produced in electron storage rings
in up to multi-GeV energy range.
AG focusing is still resorted to today, for instance in the hadrontherapy application
(Fig. 9.6), light source lattice [10], and other high energy collider design [11], as it
has the merit of compactness. On the other hand, the flexibility of separated function
optics made it more popular: it allows to introduce modular functions in complex
ring designs such as dispersion suppression sections, low-beta or insertion device
sections, long straights, et cetera. Low-emittance, high-brightness light source lattices
have complicated focusing further, by introducing longitudinal field gradient bending
systems to minimize equilibrium beam emittance [12].
Due to the necessary ramping of the field in order to maintain a constant orbit,
synchrotron accelerators are pulsed, storage rings in some cases as well, high energy
colliders in particular to bring beams to highest store energy. The acceleration is
cycled and the accelerating voltage frequency as well in ion accelerators, from injec-
tion to top energy. If the ramping uses a constant electromotive force, then (Eq. 8.3)

t
. B(t) ≈ (9.1)
τ
9.1 Introduction 327

Fig. 9.7 Cornell rapid

cycling synchrotron, 5 GeV
injector of CESR storage
ring [13]

Ḃ = d B/dt does not exceed a few Tesla/second, thus the repetition rate of the accel-
.

eration cycle is of the order of a Hertz. If instead the magnet winding is part of a
resonant circuit then the field oscillates,

B̂
. B(t) = B0 + (1 − cos ωt) (9.2)
2
so that, in the interval of half a voltage repetition period (i.e., .t : 0 → π/ω) the
field increases from an injection threshold value to a maximum value at highest
rigidity, . B(t) : B0 → B0 + B̂. The latter determines the highest achievable energy:
. Ê = pc/β = q B̂ρc/β. The repetition rate with resonant magnet cycling can reach a
few tens of Hertz, a technique known as a rapid-cycling synchrotron (RCS). In both
cases anyway B imposes its law and other parameters, comprising the acceleration
cycle, the RF frequency in particular, will follow B(t).
Instances of RCS rings include Cornell 12 GeV, 60 Hz electron AG synchrotron
[14] (Fig. 9.7), commissioned in 1967 with a 7 GeV beam, a world record at the time,
and still in operation half a century later as the injector of Cornell 5 GeV storage ring
(CESR/CHESS) [15]; Fermilab 8 GeV, 60 Hz Booster, which provides protons for
the production of neutrino beams; the 30 GeV 500 kW proton beam J-PARC facility
in Japan. Rapid cycling is also considered in ion-therapy applications (Fig. 9.6).
To conclude on these preliminaries, lets mention the giants among accelerator
facilities which nuclear (NP) and particle (HEP) physics research laboratories are:
so far, strong focusing synchrotrons happen to be the building blocks from which
328 9 Strong Focusing Synchrotron

Fig. 9.8 RHIC complex at the Brookhaven National Laboratory (left) [2], a cascade of 4 strong
focusing ion synchrotrons: the AGS and its Booster, and the 3.8 km circumference intersecting
RHIC rings, in motion towards the EIC project (right) [16] which will add 2 electron synchrotrons:
an 18 GeV storage ring and its RCS injector

they are constructed. This is so at the CERN LHC complex. This is apparent also in
Fig. 9.8 which shows RHIC heavy ion collider complex, and its planned evolution,
the Electron-Ion Collider [17]1 . The next colliders could be linacs, it was at SLAC
with the SLC [18], it was the plan with such projects as TESLA [19], the NLC [20].
The interest of NP and HEP will decide on the research tools: more large synchrotron
rings for a muon collider [21], an FCC-ee, -hh and other -eh [8], or high gradient
linacs for the ILC [22] or for ReLic .e+ e− collider [23]. Or new acceleration methods
and technologies?

9.2 Basic Concepts and Formulæ

Alternating gradient focusing is sketched in Fig. 9.9. An order of magnitude of the

focusing index can be estimated from the fields met in these structures: say a maxi-
mum B.∼1 Tesla in the dipole gap, same at pole tip in quadrupoles .∼10 cm off axis.
The latter results in . /\B
/\x
∼ 10 T/m, the former in meters to tens of meters dipole
curvature radius. All in all, in absolute value,
0∼2
ρ ∂ B 10[m]
n=−
. ∼ × 10[T/m] ∼ 101∼3 >> 1 (9.3)
B ∂x 1[T]

much greater than in a weak focusing structure, characterized by .0 < n < 1.

1 Beam polarization studies have been using zgoubi in all five EIC synchrotrons.
9.2 Basic Concepts and Formulæ 329

Fig. 9.9 Horizontally focusing lenses (field index .n >> 0, the solid red trajectory) are vertically
defocusing (.n << 0, the dashed blue trajectory), and vice versa. This imposes alternating gradients
in order for a sequence to be globally focusing, for both planes

9.2.1 Components of the Strong Focusing Optics

Combined Function (AG) Optics

This is, typically, the BNL AGS and CERN PS optics, using dipoles that ensure both
beam guiding and focusing (Fig. 9.2). Separate quadrupole and multipole lenses have
later been introduced as they provide knobs for the adjustment of optical functions
and other parameters. AG optics is still topical in modern designs, as in the iRCMS
whose six .60◦ arcs are comprised of a sequence of five focusing and defocusing
combined function dipoles [9], Fig. 9.6.

Field

Referring to normal conducting magnet technology, a hyperbolic pole profile

(Fig. 9.1) is an equipotential (a line of constant scalar potential .V ) of equation

. Vpole = A x y

at the origin of a magnetic field.B = grad V , everywhere perpendicular to the equipo-

tential. A combined function dipole with mid-plane geometrical symmetry is defined
by materializing two equipotentials, at .±Vpole (Fig. 9.10). This results in a vertical
field component . B y = ∂V /∂ y = Ax, and therefore a radial field index
|
ρ ∂ B y || ρ
n=− = A
B y ∂x |y=0
.
By

A is a constant, typically up to .∼10 T/m, cf. Eq. 9.3. The pole profile opens up
.
either inward (toward the center of curvature, a horizontally focusing dipole, verti-
cally defocusing) or outward (a vertically focusing dipole, horizontally defocusing),
Fig. 9.11.
In a bent AG dipole a line of constant field is an arc of a circle; the field guides
the reference particle along the arc in the median plane. The mid-plane field can be
expressed under the form
330 9 Strong Focusing Synchrotron

Fig. 9.10 Symmetric

materialization of pole
profiles, at .±V . Nothing
would preclude materializing
poles at .V1 and .−V2
potentials, with the same
resulting field between the
poles

Fig. 9.11 Beam focusing in

combined function dipoles.
The center of curvature is to
the left. The pole profile
follows an equipotential
. V = Ax y. Top: the pole
profile opens up towards the
center of curvature .→ the
dipole is horizontally
converging (vertically
diverging: current I comes
out of the page, force .F
results from field .B).
Bottom: pole profile closing
toward the center of
curvature .→ the dipole is
horizontally diverging,
vertically converging

( ( )2 ( )3 )
r − r0 r − r0 r − r0
. B y (r, θ) = G(r, θ) B0 1+n + n, +n ,,
+ ...
r0 r0 r0
(9.4)
with .r0 the reference (normally the orbit) radius. Higher order indices, sextupole .n , ,
octupole .n ,, , . . ., may be residual effects from fabrication tolerances, magnetic satu-
ration, deformation of yoke with years, etc., or included by design, with significant
value.
In a straight AG dipole, a line of constant field is a straight line; an instance
is the AGS main magnet (Fig. 9.2). Another instance is the Fermilab recycler arcs
permanent magnet dipole, which includes quadrupole and sextupole components [24,
25]. The modeling of the field in a straight combined function dipole can be derived
from the scalar potential of Eq. 9.5.
9.2 Basic Concepts and Formulæ 331

Separated Function Optics

In a separated function lattice quadrupole lenses ensure the essential of the focusing,
main bends have zero index. In smaller rings though, geometrical focusing in bending
magnets may be significant (Sect. 8.2.1, Fig. 8.6). Wedge angles in addition may be
introduced and contribute horizontal and vertical focusing/defocusing (Fig. 8.8).
Higher order multipole lenses are used for the compensation of adverse effects:
coupling, aberrations, space charge, impedance, etc., and for beam manipulations:
controlling the coupling, resonant extraction, etc.
The field in a multipole of order .n (.n = 1, 2, 3, etc.: dipole, quadrupole, sex-
tupole, etc.) derives, via .B = gradV , from the Laplace potential [26]
⎧ ⎫( ]
⎨E ∞
(x 2
+ y 2 q ⎬ E
)
n
x n−m y m π
q (2q)
. Vn = (n!) (−1) αn,0 (s) q
2
sin m (9.5)
⎩ 4 q!(n + q)! ⎭ m=0 m!(n − m)! 2
q=0

(2q)
where .αn,0 (s) = d 2q αn,0 (s)/ds 2q accounts for the .s-dependence of the potential.
Technologies for multipoles and combined multipoles include pole profiling, per-
manent magnets [24, 27], superconducting .cos nθ winding as in RHIC and LHC
colliders, and variants. E
In a hard-edge field model the . ∞ q=0 series is reduced to the .q = 0 term, with the
following outcomes [28, 29].

Quadrupole

The equipotential (the pole profile) is an equilateral hyperbola, of equation .Gx y =

constant in an upright quadrupole (left figure below), and .G(x 2 − y 2 ) = constant in
a .π/4 skew quadrupole (right). The resulting field writes

Upright quadrupoles are used for focusing, skew quadrupoles are used to compensate,
or introduce, transverse coupling. The focusing strength
{
1 G(s) ds
.K = (9.6)
L p/q

is momentum-dependent.
332 9 Strong Focusing Synchrotron

Sextupole

The equipotential satisfies . H (3x 2 y − y 3 ) = constant in an upright sextupole (left),

. H (x − 3x y ) = constant in a .π/6 skew sextupole (right), with resulting field
3 2

Upright sextupoles introduce a vertical field component . B y ∝ x 2 , they are used to

correct optical aberrations, to modify the momentum dependence of the wave num-
bers .νx , ν y , and in beam manipulations such as resonant extraction. Skew sextupoles
introduce a radial field component . Bx ∝ y 2 , they are used to correct optical aberra-
tions.

Octupole

The equipotential pole profile satisfies . O(x 3 y − x y 3 ) = constant in an upright

octupole (left), . O(x 4 − 6x 2 y 2 + y 4 ) = constant in a .π/8 skew octupole (right),
yielding the field

Upright octupoles are used to introduce a vertical field component . B y ∝ x 3 ; skew

octupoles introduce a vertical field component. B y ∝ y 3 . Octupoles are used to correct
aberrations, or to modify the amplitude dependence of wave numbers.

9.2.2 Transverse Motion

The transverse motion of a particle in the . S-periodic lattice of a cyclic accelerator,

at design momentum . p0 and with curvature radius .ρ0 , satisfies Hill’s equations2

2 Acceleration, or deceleration, adds a velocity term, betatron damping results. This is addressed in
“Betatron damping”, Sect. 10.2.3, where it accounts in addition for a non-constant varying orbital
radius.
9.2 Basic Concepts and Formulæ 333

d2x 1 /\p d2 y
. + K x (s)x = , + K y (s)y = 0 (9.7)
ds 2 ρ0 p0 ds 2

where . K x (s), . K y (s) have the periodicity of the lattice (. K x (s + S) = K x (s)), and
y y
depend locally on the nature of the optical elements, in the following way.
Case of ⎧
/ 1−n ( )
⎨ Kx = ρ0 ∂ B y
. − dipole :
ρ20 n = − (9.8)
/ n B0 ∂x
⎩ Ky = 2
ρ0
⎧ ( )
⎨ tan ε
Kx =± δ(s − sw ) with ε < focusing
> 0 if defocusing
. − a wedge at s = sw : ρ0
⎩ y
( )
±G 1 field at pole tip
− quadrupole : K x = ; =0 gradient G =
. Bρ ρ0 radius at pole tip
y

1
. − drift space : K x = K y = 0; =0
ρ0

By contrast with the betatron and weak focusing technologies, strong focusing
with its independent focusing (.G > 0) and defocusing (.G < 0) gradient families
allows separate adjustment of the horizontal and vertical focusing strengths, and
wave numbers as a consequence.
The on-momentum (. p = p0 ) closed orbit coincides with the reference axis of the
optical elements. The betatron motion for an on-momentum particle satisfies Eq. 9.7
with./\p = 0. Solving the latter (see section “Betatron Motion”) requires introducing
two independent solutions .u 1 (s) (Eq. 8.12), the linear combination of which yields
2
the pseudo harmonic motion (Eq. 8.14)
| ( )
| √ { ds
| u(s) = βu (s)εu /π cos + ϕ
| βu (s)
u
.| / ( ) ( ) (9.9)
| ε /π { ds { ds
| u (s) = −
, u
sin + ϕ + α(s) cos + ϕ
| βu (s) βu (s)
u
βu (s)
u

The motion satisfies the Courant-Snyder invariant, namely (Fig. 9.12)

εu
γ (s)u 2 + 2αu (s)uu , + βu (s)u ,2 =
. u (9.10)
π
i.e., the surface of the phase space ellipse is a constant of the motion. Its form
and orientation (Fig. 9.12) change along the period as a consequence of the strong
334 9 Strong Focusing Synchrotron

Fig. 9.12 Courant-Snyder

invariant and turn-by-turn
harmonic motion along the
invariant, observed at some
azimuth .s. The aspect ratio
of the ellipse depends on the
observation azimuth .s but its
area .εu is invariant

modulation of the betatron functions (Fig. 9.13), far more than in a weak focusing
lattice which features weak betatron modulation: .αu (s) ≈ 0 and .βu (s) ≈ constant
(Figs. 8.9 and 8.10).
Beam envelopes are given by the extrema,
/ /
εx εy
. x̂ env (s) = ± β x (s) , ŷenv (s) = ± β y (s) (9.11)
π π

Fig. 9.13 Optical functions around SATURNE 2 synchrotron, a 4-period FODO cell lattice
9.2 Basic Concepts and Formulæ 335

Phase Space Motion

Write the two independent solutions .u 1 (s) (Eq. 8.12) under the form
2

and u 2 (s) = u ∗1 (s) = F ∗ (s) e−iμ S

s s
u (s) = F(s) × iμ S
. 1
,,,, ,e ,, , (9.12)
μ −periodic
2πS
S−periodic

where ( )
{ s ds s
/ i 0 −μ
. F(s) = βu (s) e βu (s) S (9.13)

Introduce { s
ds s
ψu (s) =
. −μ (9.14)
0 βu (s) S
√
so that . F(s) = βu (s) eiψu (s) . Equation 9.9 thus takes the form
|
| μ −periodic
2πS
| S−periodic
| ,/ ,, ,, [ ,, ,
]
| s
| u(s) = βu (s)εu /π cos ν + ψu (s) +ϕu
.|
| R , ,, , (9.15)
| / S−per.
|
| u , (s) = − εu /π sin [ν s + ψu (s) + ϕu ] + α(s) cos [ν s + ψu (s) + ϕu ]
| βu (s) R R

Nμ
where .ν = . Thus, as the betatron function .βu (s) and phase .ψu (s) are . S-periodic,
2π
the turn-by-turn motion observed at a given azimuth .s (i.e., .u(s), u(s + S), u(s +
2S), …) is sinusoidal and its frequency is .ν = N μ/2π. Successive particle posi-
tions .(u(s), u , (s)) in phase space lie on the Courant-Snyder invariant (Eq. 9.10). The
working point .(νx , ν y ) fully characterizes the first order optical setting of the lattice.

Off-Momentum Motion

The motion of an off-momentum particle satisfies the inhomogeneous Hill’s hori-

zontal differential Eq. 9.7. The chromatic closed orbit

δp
x (s) = Dx (s)
. ch (9.16)
p

is a particular solution of the equation, its periodicity is that of the cell.

By contrast with a weak focusing lattice where chromatic closed orbits are parallel
(Eq. 8.26), in a strong focusing lattice they are distorted (Fig. 9.13), their excursion
336 9 Strong Focusing Synchrotron

depends on the distribution along the cell of (i) the dispersive elements which are
the dipoles, and (ii) the focusing.
The horizontal motion of an off-momentum particle is a superposition of the
particular solution (Eq. 9.16) and of the betatron motion, solution of the homogeneous
Hill’s equation (Eq. 9.15), namely
/ ({ )
εx ds δp
. x(s) = x β (s) + x ch (s) = βx (s) cos + ϕx + Dx (s) (9.17)
π βx p0

whereas the vertical motion is unchanged (Eq. 9.15 taken for .u(s) ≡ y(s)).

Chromaticity

The focusing strength of combined function dipoles and quadrupoles is a decreasing

function of particle rigidity . Bρ = p/q (Eq. 9.8). In a ring this affects the horizontal
and vertical wave numbers, an effect quantified as the chromaticity, .ξx,y . To the first
order in .δ p/ p, this writes
δp
.δνx,y = ξx,y (9.18)
p

A linear lattice has a natural chromaticity. Over a distance .L it is given by

{
−1
ξ
. x,y = βx,y (s)K x,y (s)ds (9.19)
4π L
{
Use a circular integral, . in the case of a ring. The natural chromaticity is a negative
quantity: focusing decreases with increasing momentum.
One consequence of the chromaticity is that beam momentum spread.δ p/ p results
in a tune spread .δνx,y = ξx,y × δ p/ p, a beam occupies an extended area in the tune
diagram. For this reason in particular, the chromaticity is usually corrected. This is
realized by placing sextupoles in dispersive sections, at least two families: a family
of horizontal lenses (strength . Hx ) located at large .βx and a family of vertical lenses
(strength . Hy ) located at large .β y .
The effect leaned on is the following:
– betatron motion .xβ (s) of particles with momentum . p0 + /\p is around an off-
centered, chromatic closed orbit .xch (s) (Eq. 9.16);
– introducing a sextupole results | in a local gradient as . B y ∝ (xch + xβ )2 = xch
2
+
∂ B | /\p
2xch xβ + xβ2 , namely, . ∂x |y
= 2xch = 2Dx p . This results in a focusing force
x=xch
proportional to .δ p/ p. Sextupoles contribute to chromaticity (or its compensation)
following {
1
.ξx,y = Hx,y (s)βx,y (s)Dx (s)ds (9.20)
4π
9.2 Basic Concepts and Formulæ 337

9.2.3 Resonances

Consider the excitation of transverse beam motion by a generator of frequency .g

located at some azimuth along the ring [29]. The action of the excitation . S × sin gt
on the oscillating motion .u(t) can be written under the form

d 2u
. + ω 2 u = S sin gt (9.21)
dt 2
Assume harmonic motion for simplicity (as in a weak focusing lattice). Take gen-
erator amplitude . S = constant, the solution (superposition of the solution of the
homogeneous differential equation and of a particular solution of the inhomoge-
neous differential equation) writes
S
.u(t) = U cos(ωt + ϕu ) + sin gt (9.22)
ω 2 − g2

If betatron motion and excitation are in synchronism, i.e. on the resonance, .ω = g,

a particular solution of Eq. 9.21 is

the amplitude of the oscillatory motion grows rapidly with time, at a rate .|St/2g|.
Assume the amplitude. S to be.T , -periodic instead, angular frequency.ω , = 2π/T , ,
take its Fourier expansion
∞
E
. S(t) = a p cos( pω , t + ϕ p )
p=0

the equation of motion thus writes

E ∞
d2u
2
+ ω2 u = a p cos( pω , t + ϕ p ) sin gt =
dt
p=0
E ap [
. ∞
] (9.23)
sin[(g − pω , )t + ϕ p ] + sin[(g + pω , )t + ϕ p ]
2
p=0

Resonance may occur at generator frequencies .g = ω ± pω , , the strength depends

on the amplitude .a p of the excitation harmonics. A generator at some point in the
lattice excites all harmonics with equal amplitudes .a p . In the case of an extended
excitation source, low harmonics only matter.
338 9 Strong Focusing Synchrotron

Sextupole and Octupole Resonances

The horizontal motion in the presence of sextupoles (. B y (θ)|y=0 = S(θ)x 2 ) satisfies

d2x
. + νx2 x = S(θ)x 2 (9.24)
dθ2

Assume weak perturbation of the motion, so that .x(θ) ≈ x̂ cos(νx θ + ϕx ), the solu-
E Assume also . S(θ) .2π-periodic. Substitute its Fourier
tion for unperturbed motion.
series expansion . S(θ) = ∞p=0 a p cos( pθ + ϕ p ) in Eq. 9.24, develop to get

[
d2x x̂ 2 E
∞
+ ν 2
x x = a p cos( pθ + ϕ p ) +
dθ2 2 p=0
. ]
1 E
∞ [ ]
2
a p cos[( p − 2νx )θ + ϕ p − 2ϕx ] + cos[( p + 2νx )θ + ϕ p + 2ϕx ]
p=0
(9.25)
Thus resonance may occur at the betatron frequency families .νx = ± p, .νx = ±( p −
2νx ), and .νx = ±( p + 2νx ), i.e.,
[
νx = p
.
3νx = p

In the case of a single sextupole in the ring, all the harmonics . p are excited with the
same amplitude .a p .
An octupole introduces a field component . B y (θ)|y=0 = O(θ)x 3 . A similar devel-
opment yields
⎡
νx = p
. | 2νx = p
4νx = p

Resonances in a general manner occur at betatron frequencies satisfying

.mνx + nν y = integer

In this coupling regime one has

εx εy
. − = constant, an invariant of the motion (9.26)
m n
From this it results that,

– if .m and .n have opposite signs the resonance causes energy exchange between the
εx ε
horizontal and vertical motions: . |m| + |n|y = constant, an increase of .εx correlates
9.2 Basic Concepts and Formulæ 339

with a decrease of.ε y and vice-versa. In the presence of linear coupling for instance,
ν − ν y = integer, .εx + ε y = constant. An increase in motion amplitude anyway
. x
may cause particle loss, an issue in cyclotrons where the Walkinshaw resonance
.ν x = 2ν y causes vertical beam loss due to the increase of .ε y ;

– if .m and .n have the same sign the resonance is liable to induce motion instability:
εx ε
.
m
− ny = constant, .εx and .ε y may both increase with no limit.

Resonant Extraction

Resonant extraction is based on the effect of a non-linear force on a dynamical sys-

tem. A linear regime, under the effect of linear forces, satisfies Eq. 9.7. If .x(s) is a
stable solution, so is .λx(s) (.λ a proportionality constant). Introducing a non-linear
force modifies the equation of motion, into for instance

d2x
0.
. + K x (s)x = S(s)x 2 : sextupole perturbation,
ds 2
d2x
.0 . + K x (s)x = O(s)x 3 : octupole perturbation,
ds 2
If .x(s) is a stable solution, it may no longer be the case for .λx(s). If .x(s) is small
enough the motion, subject to linear and non-linear forces, is quasi-linear and sta-
ble. However, increasing the motion amplitude will at some point result in unstable
motion. In the .(x, x , ) phase space, the stable regime is bounded by a separatrix.
Outside the latter the motion is essentially unstable, or liable to reach amplitudes
beyond transverse acceptance of the accelerator (Fig. 9.14).

Fig. 9.14 Horizontal motion

near a 3rd integer resonance.
Within the triangle separatrix
the motion is stable. Outside
the triangle, motion reaches
large amplitudes. An
electrostatic septum extracts
particles which jump to the
right of the septum (into the
extraction channel) during
their motion
340 9 Strong Focusing Synchrotron

Fig. 9.15 In the presence of

RF, particles oscillate in the
vicinity of the synchronous
phase. Above transition, in
this schematic

9.2.4 Acceleration. Synchrotron Motion

Particle motion in longitudinal phase space .(phase, momentum) and its stability
are determined by the lattice and by the acceleration parameters, as introduced in
Sect. 8.2.2. They include the

– RF . f rf = h f rev , {
– voltage .V (t) = V̂ sin ωrf dt,
– synchronous phase .φs (phase of the particle in synchronism with the RF oscilla-
tion), which increases√ by .2πh per turn,
– transition .γtr = 1/ α (Fig. 8.15).

In the case of weakly modulated betatron functions (weak focusing lattice; AG

lattice to some extent), .α ≈ 1/νx2 so that

γ ≈ νx
. tr

This is the case of SATURNE 1: a weak focusing lattice (see Chap. 8 and simulation
exercises there) operated above transition as .γtr = νx ≈ 0.6. In the AGS at BNL the
working point is .νx ≈ 8.7 whereas .γtr = 8.4 ≈ νx ; transition is crossed as proton
beams are accelerated from .γ ≈ 3 to .γ ≈ 25. Instead, SATURNE 2 strong focusing
lattice was operated at negative .α, .η = γ12 − α does not cancel, .γtr is pure imaginary.
The energy gain per turn at the cavity is

/\W = 2π R qρ Ḃ = q V̂ sin φs
.

/\W is imposed by the field law in order to ensure that at all time the synchronous
.
particle momentum satisfies

. ps (t) = q B(t)ρ
9.2 Basic Concepts and Formulæ 341

Phase Stability

Particles with phase and momentum offsets .(/\φ, /\p/ p) = (φ − φs , ( p − ps )/ ps )

in the vicinity of the synchronous particle at .(φs , ps ) undergo periodic longitudinal
oscillations (Fig. 9.15). The longitudinal motion satisfies the differential equations

d/\φ /\p d(/\p/ p) e V̂ ωs

. = hηωs , = [sin φ − sin φs ] (9.27)
dt p dt 2πβs2 E s

If peak amplitudes are small the differential Eq. 9.27 yield

d 2 /\φ
. + g2s /\φ = 0 (9.28)
dt 2
the motion is sinusoidal, with a synchrotron angular frequency
/
c ηhq V̂ cos φs
.gs = (9.29)
R 2π E s

The synchrotron tune, number of synchrotron oscillations per revolution, writes

/
gs 1 ηhq V̂ cos φs
.νs = = (9.30)
ωrev βs 2π E s

Synchrotron oscillations are slow compared to betatron oscillations, typically .νs ∼

νx,y /102∼3 . Motion stability requires .g2s > 0, or

η cos φs > 0
.

Longitudinal motion in .(φ, φ̇/gs ) phase space is on a circle. The extent in phase and
energy, or momentum, of the small amplitude oscillations satisfy

^ ^
^ = hη E s /\E = hη E s β 2 /\p
/\φ
. (9.31)
s
ps Rgs E s ps Rgs p

The bunch length is

R^
. L bunch = /\φ (9.32)
h

Separatrix

If peak amplitudes are large the oscillations are non-linear and, assuming slow accel-
eration
342 9 Strong Focusing Synchrotron

Fig. 9.16 Longitudinal motion separatrix in .(φ, dp/ p) phase space, and some stable as well as
unbounded motions. Case of SATURNE 2 at injection energy, 50 MeV. From left to right: case
of .φs = 0 (stationary bucket), .φs = 15, 30, and .60◦ . Small motions are centered on .φs , their
synchrotron tunes satisfy Eq. 9.30. The momentum acceptance (height of the separatrix) satisfies
^
Eq. 9.36, with respectively .± /\p
p ≈ 0.00496, 0.00392, 0.00290 and .0.00107

d 2 /\φ sin φ − sin φs

. + g2s =0 (9.33)
dt 2 cos φs

A first integral of this equation is the equation of the separatrix (Fig. 9.16)

φ̇2 cos φ + φ sin φs cos(π − φs ) + (π − φs ) sin φs

. − g2s = −g2s (9.34)
2 cos φs cos φs

This defines two locations where .φ̇ changes sign, i.e. .φ̇ = 0, namely,
(i) .φ1 = π − φs ,
(ii) .φ2 such that .cos φ2 + φ2 sin φs = cos(π − φs ) + (π − φs ) sin φs .
The motion is stable, oscillatory, within the domain .φ e [φ1 , φ2 ], the “bucket”,
and unbounded beyond. The bucket height is obtained for .φ = φs , namely, from
Eq. 9.34
φ̇max /
. = 2 [2 − (π − 2φs ) tan φs ] (9.35)
gs

Expressed in momentum,
/
^
/\p 1 q V̂
.± =± [2 cos φs − (π − 2φs ) sin φs ] (9.36)
p βs πhη E s

Its dependence on.φs is represented in Fig. 9.17. Stationary bucket mode, i.e..sin φs =
0, has greatest acceptance. The latter decreases in accelerated bucket mode as .φs →
π/2 (Fig. 9.16).

Adiabatic Damping of Synchrotron Oscillations

The equation of motion, Eq. 9.33, assumes a slow acceleration rate, .dTrev /dt << 1,
such that . ps (t), .η, possibly .V̂ , and thus .gs change slowly during synchrotron
9.2 Basic Concepts and Formulæ 343

Fig. 9.17 Dependence of

the momentum
( extent of the
bucket . normalized to
/
1 q V̂ )
.
βs πhη E s on the
synchronous phase √.φs . It
takes its value in . 2 → 0
for .sin φs : 0 → 1

oscillations and therefore can be considered constant. The extreme phase and momen-
tum excursions during acceleration satisfy
( )1/4
^ ∝ η
/\φ
R 2(γ V̂ cos φs )
. 1/4 (9.37)
^
/\p 1 V̂ cos φs
∝
p βs ηγ 3 R 2

In the case of acceleration on a fixed orbit (constant radius . R),

.^ × /\p
/\φ ^ = constant (9.38)

Adiabatic Damping of the Betatron Oscillations

The mechanism is described in Sect. 8.2.2 (Fig. 8.14), the equations of motion are
addressed in Sect. 10.2.3. In the case of an adiabatic change of momentum . p =
βγm 0 c (a slow change compared to the betatron motion oscillation frequency) the
transverse motion damping satisfies

. p εu = constant, or βγεu = constant (9.39)

Coordinate damping satisfies (Eq. 10.22 with orbit radius . R = constant)

√ √
. x, y ∝ 1/ p, x , , y , ∝ 1/ p. (9.40)

9.2.5 Synchrotron Radiation, Dynamical Effects

Emittance growth upon SR matters in high.γ rings, electron rings so far, muon collider
possibly in the future [30] and other FCC lepton and hadron collider [8].
344 9 Strong Focusing Synchrotron

The stochastic nature of SR and the energy loss it results in, have been introduced
in Chap. 5. Dynamical effects in a synchrotron ring are further addressed here [31,
32].

Motion Invariants

In the absence of perturbation by synchrotron radiation, particle motion satisfies the

Courant-Snyder (Eq. 9.41) and longitudinal (Eq. 9.42) phase-space invariants

ε = γu (s)u 2 + 2αu (s)uu , + βu (s)u ,2 (u = x or y)

. u (9.41)
( )2
αE s δ^E
.εl = (9.42)
2 gs Es

Under the effect of stochastic SR, individual invariants can in general not be deter-
mined, averages over particle ensembles are considered instead (noted .(∗) in the
following), they evolve according to

dεu εu
. = − + Cu (9.43)
dt τu

towards a stationary solution

ε
. u,eq = C u τu (9.44)

where .Cu is a constant at fixed energy (storage ring), with characteristic time

Tr ev E s
τ =
. u (9.45)
Us Ju

J are the partition numbers (lattice properties), respectively horizontal, vertical,

. n=x,y,l

longitudinal,
. Jx = 1 − D, Jy = 1, Jl = 2 + D (9.46)

where

Dx (1 − 2n)/ρ3
D=
.
1/ρ2
{
In this expression, .(∗) = 2π1R dipoles (∗)ds, .n is the field index—case of combined
.
function dipoles, . Dx is the dispersion function, The partition numbers satisfy the
Robinson theorem
. Jx + J y + Jl = 4 (9.47)
9.2 Basic Concepts and Formulæ 345

Table 9.1 Common expressions for the energy loss per turn, .Us (E-loss), for the damping times
and equilibrium emittances, in the hypothesis of an isomagnetic lattice. Their scaling with.γ is given
in the 2nd row
E-loss .εl,eq .σl .τl .εx,eq a .τx .ε y,eq .τ y

Scaling: .γ 4 .γ 3/2 .1/γ 1/2 .1/γ 3 .γ 2 .1/γ 3 .1/γ 3

αE s Cq γ . αc σ Cq γ 2
2
E s4 T E T E T E
.Cγ ρ .
gs Jl ρ gs /\E . rev s
Us Jl .
Jx ρ H . rev s
U J .<< εx . rev s
Us J y
{ [ E
] s x
aH
. = 1
L dip
ds
dip βx Dx2 + (αx Dx + βx Dx, )2 , integral over the dipoles

Common expressions for the calculation of the energy loss and equilibrium quan-
tities, in the hypothesis of an isomagnetic lattice, are recalled in Table 9.1.
Vertical emittance results from coupling, always present in a ring, due for instance
to a loss of median plane symmetry, or to fringe fields, or excited on purpose to control
the vertical emittance as in light sources. Given the coupling factor .κ—normally
.<0.1, the vertical and horizontal emittances satisfy

e = κex ,
. y ex + e y = e0 (9.48)

where.e0 is the equilibrium horizontal emittance in the absence of coupling (Table 9.1).
The basic considerations above hold for a defect-free planar ring. Things can be
(as usual) more complicated, for instance in the presence of vertical dispersion.

Field Scaling

Particle stiffness decrease upon SR loss causes these to experience increased field
strength (.1/ρ in dipoles,.G/Bρ in quadrupoles, etc.). In the case of beam lines (which
may include high energy ERLs [11]), this effect may be taken care of by scaling the
magnetic fields to the theoretical average energy loss (Eq. 5.12), namely
E 2
/\E scaling =
. r0 ecγ 3 B/\θ (9.49)
bends
3

In a storage ring the energy lost by SR is restored by the RF system, bends and lenses
are operated at constant field. In pulsed regime such as in a booster injector, bends
and lenses are operated at constant strength during acceleration.

9.2.6 Visible Synchrotron Radiation. Interference

Visible SR was first observed at the GEC 70 MeV. For this reason it has been
introduced in the Weak Focusing Synchrotron chapter, Sect. 8.2.3. The SR spectrum
at that energy peaks—has its critical frequency—in the visible region. The matter
346 9 Strong Focusing Synchrotron

Fig. 9.18 Left: typical shape of the . E σ (τ ) and . E π (t) electric impulse components of the Poynting
vector, emitted by a 2.5 GeV electron on a.ρ = 53.6 m circular trajectory in a.l = 20 cm-long dipole,
as observed in the laboratory.. E σ, π (τ ) are obtained from the stepwise integration of electron motion
through the magnet, which provides the ingredients to compute Eq. 8.36, accounting for the retarded
time .t = τ − r (t)/c (Eq. 8.37). Right: the spectral brightness of the .σ component of the radiation
allows comfortable beam diagnostics conditions in the visible range (.hω ∼ 0.5 eV)

is developed further in the present chapter, in regard with the use of visible SR for
beam diagnostics in electron and high energy proton rings [31, 33].
An example of the use of visible SR from a proton beam is found at the CERN SPS,
where edge radiation was used at 270 GeV for beam imaging [34]. At that energy
in the SPS, the critical frequency (the peak brightness) is in the infrared region.
Undulator radiation, more intense, was used down to 200 GeV [35], in the .p − p
collider era (1980s). Another example is the LHC synchrotron light profile monitor,
a major beam monitoring tool at injection energy, 450 GeV [36], [37, Appendix C].
An example of the use of visible SR from a high energy electron beam is found
at the former LEP, where it was produced in a dedicated 4-dipole miniwiggler. The
critical frequency in a high energy electron ring is way above the visible range.
In such case, visible SR can be dealt with in terms of low-frequency SR [38], a
method which can be extended to the analytical treatment of SR interference [37].
The underlying theoretical material is recalled here. It is resorted to in the exercises,
to cross check Poynting computation from raytracing (using Eq. 8.36).

Low Frequency SR

A typical electric field impulse from a LEP miniwiggler dipole, and the resulting
spectral brightness, as observed in the laboratory, are displayed in Fig. 9.18. The
LEP 4-dipole miniwiggler was subject to visible light interference from 4 coherent
sources, the effect is illustrated in Fig. 9.19.
A doublet of LEP miniwiggler dipoles, in both cases of same sign and opposite
sign dipoles, is the object of numerical simulations in Exercise 9.6. It is on the
other hand treated theoretically in [37, Sect. 3.1]. The latter provides all necessary
9.2 Basic Concepts and Formulæ 347

Fig. 9.19 An interferencial spectrum, case of LEP 4-dipole miniwiggler [39]. By contrast with the
single dipole case (Fig. 9.18), the spectral brightness of the .σ component cancels in the low energy
end of the spectrum

material for cross checks of numerical outcomes from the stepwise integration of
electron motion.

9.2.7 Polarization, Resonances

In a weak focusing optics lattice, radial field components experienced by a particle in

the course of its vertical betatron motion are small, which results in weak depolarizing
resonances (Sect. 8.2.4). By contrast, strong focusing field gradients in the combined
function dipoles and/or focusing lenses of strong focusing optics results in strong
radial field components and therefore strong depolarizing resonances.
Spin precession and resonant spin motion in the magnetic components of a cyclic
accelerator have been introduced in Sects. 3.2.5 and 4.2.5. The general conditions
for depolarizing resonance to occur have been introduced in Sect. 8.2.4. In a strong
focusing synchrotron they essentially result from the radial field components in the
focusing magnets and their strength is determined by the lattice optics, as follows.

Strength of Imperfection Resonances

Imperfection, or integer, depolarizing resonances are driven by a non-vanishing ver-

tical closed orbit . yco (θ) which causes spins to experience periodic radial fields in
focusing magnets, dipoles in combined function lattices and quadrupoles in sepa-
rated function lattices, namely,

. Bx (θ) = G y(θ) = K (θ) × B0 ρ0 × yco (θ) (9.50)

with .θ the orbital angle and . B0 ρ0 the lattice rigidity. Resonance occurs if the
spin undergoes an integer number of precessions over a turn: it then experiences
348 9 Strong Focusing Synchrotron

1-turn-periodic torques, which cause it to move away from the stable .n0 direction as
field perturbations along the closed orbit add up coherently. Thus resonances occur
at integer values

. Gγn = n

A Fourier development of these perturbative fields yields the strength of the .Gγn
.

harmonic [40, Sect. 2.3.5.1]

{
R
.en
imp
= (1 + Gγ) K (θ) yco (θ) e− j Gγ(θ − α) e jnθ dθ
2π

In the thin-lens approximation, near the resonance where .Gγ − n → 0, this simpli-
fies into a series over the quadrupole fields,

1 + Gγn E
eimp =
. n [cos Gγn αi + sin Gγn αi ] (K L)i yco (θi ) (9.51)
2π Qpoles

with .θi the quadrupole location, .(K L)i the integrated strength (slice the dipoles as
necessary in an AG lattice for this series to converge) and .αi the cumulated orbit
deviation.
Orbit harmonics near the betatron tune (.n = Gγn ≈ ν y ) excite strong resonances.
Imperfection resonance strength is further amplified in P-superperiodic rings, with
m-cell superperiods, if the betatron tune .ν y ≈ integer × m × P [41, Chap. 3-I].

Strength of Intrinsic Resonances

Intrinsic depolarizing resonances are driven by betatron motion, which causes spins
to experience strong radial field components in quadrupoles, namely

. Bx (θ) = G y(θ) = K (θ) × B0 ρ0 × yβ (θ) (9.52)

The effect of resonances on spin depends upon betatron amplitude and phase, their
effect on beam polarization depends on beam emittance. Longitudinal fields from
dipole ends are usually weak by comparison and ignored. The location of intrinsic
resonances depends on betatron tune, it is given in an M-periodic structure by

. Gγn = n M ± ν y

A Fourier development of the perturbative fields yields the two families of strengths
[40, Sect. 2.3.5.2]
/ ({ )
{ s(θ)
λ x ρ0 2π εy ± j 0 ds
− νy θ
intr ±
.en = K (θ) β y (θ) e βy
e− j Gγ(θ − α(θ)) e jnθ dθ
4π 0 π
9.3 Exercises 349

In the thin-lens approximation, near the resonance where .Gγ ± ν y − n → 0, this

simplifies into a series over the quadrupole fields,
{ ] { ] /
Re(eintr
±
)+ 1 + Gγn E cos(Gγn αi ± ϕi ) + εy
.
n
intr ±
= (K L)i β y,i
j Im(en ) 4π j sin(Gγn αi ± ϕi ) π
Qpoles
(9.53)

Spin Diffusion

Spin diffusion stems from the stochastic emission of photons in magnetic fields
(Sect. 5.2.3). A change .δ in the energy offset ./\E of a particle, due to the emission of
a photon, causes a change .∂n/∂δ of the local spin precession direction. In dispersive
sections it also causes a change in the horizontal invariant, .∂ex /∂δ, and in vertical
invariant as well, .∂e y /∂δ in the presence of vertical dispersion, which in turn result
in perturbations .∂n/∂ex,y .
As far as numerical integration is concerned, spin diffusion is a sub-product of the
stepwise integration of Thomas-BMT equation (Sect. 3.2.5), and of the simulation of
stochastic emission of photons (Sect. 5.2.3). It is at work in Cornell RCS simulation,
Exercise 9.4.

9.3 Exercises

In complement to the present exercises, a tutorial on depolarizing resonances in

a strong focusing synchrotron can be found in [40, Chap. 14]. Proton, helion and
electron beams are considered, using the lattice of the AGS Booster at BNL. The
simulations explore methods for preservation of polarization, including tune-jump
quadrupoles, a solenoid, Siberian snakes, spin rotators in the case of electrons, includ-
ing synchrotron radiation and effects on polarization life time.
Note: input data files for these simulations are available in zgoubi sourceforge
repository at https://sourceforge.net/p/zgoubi/code/HEAD/tree/branches/exemples/
book/zgoubiMaterial/synchrotron_strongFocusing/.
9.1 Construct SATURNE 2 Synchrotron
Solution 9.1.
Over the years 1978–1997 the 3 GeV synchrotron SATURNE 2 at Saclay (Figs. 9.3
and 9.20) delivered polarized proton beams, and polarized deuteron and .6 Li beams
up to 1.1 GeV/nucleon, for intermediate energy nuclear physics research, including
meson production [42, 43, 45]. The separated function synchrotron was designed
ab initio for the acceleration of polarized ion beams [44], and the first strong focus-
ing synchrotron to do so—ZGS, first to accelerate polarized beams, protons and
deuterons, was a weak focusing synchrotron (Chap. 8).
350 9 Strong Focusing Synchrotron

SPES IV

Fig. 9.20 SATURNE 2 synchrotron and its experimental areas, including mass spectrometers
SPES I to SPES IV, a typical nuclear physics accelerator facility. The polarized ion sources Dioné
and Hypérion are at the top left, followed by a 20 MeV linac. In the early 1980s a synchrotron
booster, MIMAS, was added for higher polarized ion performance

SATURNE 2 is a FODO lattice with missing dipole. Its parameters are given in
Table 9.2.
(a) Simulate the main dipole using BEND. Dipole fringe fields matter in this small
ring, take them into account assuming .λ = 8 cm extent and the following Enge coef-
ficient values (Eq. 14.11, Sect. 14.3.3):

. C0 = 0.2401, C1 = 1.8639, C2 = −0.5572, C3 = 0.3904, C4 = C5 = 0

Produce the transport matrix of the dipole, check against theory. Compare with the
matrix of the hard edge model.
Produce a graph of the field across the dipole, in the median plane and at 5 cm ver-
tical distance. OPTIONS[CONSTY.=ON] can be used to force a particle to constant
Y and Z.
Simulate the F and D quadrupoles, using respectively QUADRUPOLE and MUL-
TIPOL. Compare matrices with theory.
Construct the cell. Produce machine parameters (tunes, chromaticities), check
against data, Table 9.2.
Construct the 4-cell ring. Produce a graph of the optical functions. Produce the
beam matrix.
9.3 Exercises 351

Table 9.2 Parameters of SATURNE 2 separated function FODO lattice. .ρ0 is the radius of the
reference orbit in the main dipole
Orbit length, .C m 105.5556
Average radius, . R = C/2π m 16.8
Straight sections, length:
– Short m 0.716256
– Long m 3.92148
Dipole:
– Bend angle, .α deg 22.5
– Magnetic radius, .ρ0 m 6.3381
– Wedge angle, .ε deg 2.45
Quadrupole:
– Gradient range T/m 0.5–10.56
– Magnetic length F/D m 0.46723/0.486273
Wave numbers, typical, .νx ; .ν y 3.64; 3.60
Chromaticities, .ξx ; .ξ y Negative, a few units
Momentum compaction .α 0.015
Injection energy (proton) MeV 20
Top energy GeV 3
. Ḃ T/s 4.2
Synchronous energy gain keV/turn 1.160
RF harmonic 3

(b) Accelerate a bunch comprised of a few tens of particles with Gaussian density
distributions (it can be defined using MCOBJET), from injection to top energy,
50 MeV to 3 GeV. Use harmonic 3 RF frequency, take a (unrealistic, for a reduced
number of turns) peak RF voltage .V̂ = 1 MV, and synchronous phase .φs = 30◦ .
Produce a graph of Y, Z and dp/p versus turn. Check the transverse damping
against theory.
(c) Determine the momentum acceptance of the ring at 50 MeV, with.V̂ = 10 kV peak
voltage, in the following four cases: stationary bucket (synchronous phase .φs = 0)
and accelerated buckets with .φs = 15, 30, and 60◦ .
Reproduce the longitudinal phase space graphs displayed in Fig. 9.16.

9.2 Non-Linear Motion in SATURNE 2

Solution 9.2.

(a) Simulate horizontal particle motion near a third integer resonance. Provide a
graph of the transverse phase space.
(b) Simulate horizontal particle motion near a quarter integer resonance. Provide a
graph of the transverse phase space.
352 9 Strong Focusing Synchrotron

9.3 SVD Orbit Correction

Solution 9.3.
Using SATURNE 2 ring, inject dipole defects and use SVDOC to find the corrected
orbit.
It can be done in the following way:
– place a horizontal pickup (HPU), a dipole defect (HDEF, using a thin-lens MUL-
TIPOL, length e.g. 1e-3 cm) and a dipole corrector (HKIC, using a thin-lens MUL-
TIPOL) in the middle of the QF quadrupole of the FODO cells,
– in a similar manner, place a VPU, a VDEF and a VKIC just upstream of the FODO
cell QD,
– excite V and H closed orbits by injecting random defects in HDEF and VDEF,
using ERRORS.
Use SVDOC to find the orbit correction.
Provide a graph of the orbit at the PUs, before and after correction.
In the previous setting, there is 24 defects (12 H and 12 V) and 24 correctors (12 H
and 12 V). Repeat for 24 defects and only 12 correctors per plane.
9.4 Cornell Electron RCS. Radiative Energy Loss
Solution 9.4.
Note: details regarding these simulations and their solutions can be found in the Tech.
Note EIC/57;BNL-114452-2017-IR [46].

The goal in this exercise is to simulate Cornell RCS lattice and accelerate beam,
first without synchrotron radiation, then taking it into account. In a fourth step electron
spin is added and polarization transmission through the acceleration cycle assessed.
(a) Details of the RCS geometry and lattice can be found in Ref. [14], however
a simplified 6-superperiodic version of the ring is considered here, with six
identical long straights and six identical arcs. The RCS parameters are given in
Table 9.3. The input data files are given in
– Tables 9.4 and 9.5: definition of the focusing and defocusing bends, and of
the focusing and defocusing doublets;
– Table 9.6: definition of a FODO cell;
– Table 9.7: definition of a supercell;
– Table 9.8: definition of the 6-supercell ring.
Produce the optical parameters of the ring. A TWISS command can be used for
that. Produce graphs of the closed orbit and optical functions around the ring.
(b) Raytrace a few tens of particles over 2300 turns around the ring, from 320 MeV
to 8 GeV about, ignoring radiative energy loss. Assume normalized emit-
tances .εx = ε y = 25 πµm, Gaussian densities, initial rms .δ p/ p = 5 × 10−3 .
Use CAVITE[IOPT.=3] for acceleration. Produce a graph of the three phase
spaces produce graphs of transverse and longitudinal excursions versus turn
number, check damping again expectations.
9.3 Exercises 353

Table 9.3 Cornell RCS parameters in the present simplified lattice simulation
Top energy GeV .7

Injection energy MeV 320

Circumference, simplified m 786.947
6-supercell case
Bunch
.εx , .ε y at injection .πµm .25

Bunch length mm .6

dE/E at injection .5 × 10
−3

Combined function lattice .48×FFDD

Nb of F and D cell dipoles 192

.ρ F , .ρ D m .≈95, 92

Field at 7 GeV T 0.25

Max. .βx , .β y m 29, 26
.νx , .ν y , natural 9.62, 13.82
.ξx , .ξ y , natural .−13, .−16

RF, synch. radiation

Repetition rate Hz up to 60
Acceleration rate MV/turn 3
E-loss per turn at 5, 10 GeV MeV 0.6, 9
.τx (≈ 3 ) at 5, 10 GeV
2.5
E
ms 16, 2

(c) Re-do (b) with synchrotron radiation energy loss, following SR loss theoreti-
cal material introduced in the “Betatron” Chap. 5. Use SRLOSS for radiation,
and CAVITE[IOPT.=11, Facility.=CornellSynch, .U00 = 9.48145321 × 10−6 ]
for acceleration. Check equilibrium emittances.
(d) Produce a graph of the average bunch polarization over the acceleration cycle
in (c), starting with all spins up at injection energy. Check against the resonance
spectrum over .aγ : 0.7 → 18.
9.5 Coupling in a Light Source Storage Ring
Solution 9.5.
In this exercise, it is proposed to reproduce SR damping simulations, in a case of
coupled light source lattice, detailed in JINST article [48].
Simulation of radiation damping in rings, using stepwise ray-tracing methods
(the original (1990s) ESRF lattice is concerned—today’s ESRF lattice is completely
different, minimal emittance, un-isomagnetic).
An input data file for the early ESRF lattice can be found at https://sourceforge.net/
p/zgoubi/code/HEAD/tree/branches/exemples/SRDamping/ESRFRing/coupled. It
accounts for .κ = 0.58 optical coupling, by a single skew quadrupole placed at the
begining of the lattice.
Reproduce the numerical results for this coupled case, as detailed in Sect. 5 of
that JINST article [48].
354 9 Strong Focusing Synchrotron

Table 9.4 Simulation input data files for the focusing (left) and defocusing (right) com-
bined function dipoles. They define the segments, respectively, F_BEND_S:F_BEND_E and
D_BEND_S:D_BEND_E, for use by INCLUDE commands in further input data files. These files
can be run as is: FIT will center the closed orbit across the magnet, accounting for the field scaling
by the ad hoc coefficient under SCALING

Table 9.5 Definition of focusing (left) and defocusing (right) doublets, for use by further INCLUDE
commands
9.3 Exercises 355

Table 9.6 Simulation input data file for a FODO cell

9.6 SR Electric Impulse and Interference in a Miniwiggler

Solution 9.6.
In this exercise, the electric field component of synchrotron radiation in short dipoles
is produced. An interferential spectrum is prodcued from a pair of dipoles. This
exercise is based on the LEP miniwigller configuration [37].

(a) Produce the input data file for the simulation of an electron trajectory in one
of the LEP miniwiggler dipoles schemed in Fig. 9.21. Dipole length is . L =
52.602 cm, bend angle .0.8 mrad. Electron energy is . E = 45 GeV. Produce the
electric field impulse observed at long distance in the direction .φ = ψ = 0.
Produce its spectrum.
Check the various quantities: duration of the electric field impulse, critical fre-
quency of the spectrum, etc.
(b) Consider the dipole pair of Fig. 9.21. Take distance between dipoles
.d = 23.098 m. Produce the electric field impulse observed at long distance in

the direction .φ = ψ = 0. Produce its spectrum.

Check the various quantites: duration of the electric field impulse, critical fre-
quency of the spectrum.
Repeat, in the direction .φ = 0, ψ = 0.2 mrad.
(c) Repeat (b), for the dipole pair disposed as in Fig. 9.21 [37, Sect. A].
(d) Repeat (c) for the configuration of Fig. 9.22, a case of edge radiation interfer-
ence [37, Sect. B].

9.7 Depolarizing Resonances in SATURNE 2

Solution 9.7.
Unexpectedly as it is not a systematic resonance, .Gγ = 7 − ν y was found to be
harmful to beam polarization. Produce a crossing of that resonance, for a few particles
with different momenta, and vertical invariant .ε Z ≈ 10π µm. Take peak voltage 6 kV
and synchronous phase .φs = 0.2363176 rad.
The input data file given in Table 9.14, an outcome of Exercise 9.15, can be used
as a starting point for this simulation.
356 9 Strong Focusing Synchrotron

Table 9.7 Simulation input data file for a supercell

9.3 Exercises 357

Table 9.8 Simulation input data file for Cornell RCS ring

Fig. 9.21 Synchrotron radiation electric field impulse from a pair of dipoles is observed in the
direction .(φ, ψ), with .φ the bend plane angle as shown, and .ψ the angle to the bend plane. This
schematic defines the observation direction .φ = 0
Fig. 9.22 Both dipoles have
same sign. This schematic
defines the observation
direction .φ = 0

9.8 Ion and Electron Polarization. Preservation of Polarization

More simulations regarding

– spin polarized ions and special devices and methods for the preservation of polar-
ization during acceleration, including tune jump, partial and full Siberian snakes,
etc.,
– electron spin diffusion in a storage ring and its suppression, spin matching, polar-
ization lifetime, etc.,
can be found, with complete solutions, in the USPAS Summer 2021 Spin Class
Lectures, “Polarized Beam Dynamics and Instrumentation in Particle Accelera-
tors” [47, Chap. 14].
358 9 Strong Focusing Synchrotron

Table 9.9 Simulation input data file saturneBEND.inc: computes the transport matrix of SAT-
URNE 2 dipole. It does so via parameter adjustment resorting to FIT, by imposing symmetric posi-
tioning and null in and out reference orbit coordinates. This file defines the SATURNE 2 sequence
segment satBEND_S to satBEND_E for use in subsequent INCLUDE statements

9.4 Solutions of Exercises of This Chapter: Strong Focusing

Synchrotron

9.1 Construct SATURNE 2 Synchrotron

(a) A model of SATURNE 2 synchrotron.
.• Simulation of the main dipole, using BEND, is given in Table 9.9, the simulation
includes finding the reference orbit using FIT, and delivering the transport matrix.
The theoretical transport matrix, Eq. 14.7, can be used to check the latter.
Looking up the execution listing zgoubi.res, one finds parameters used for the
stepwise raytracing through BEND (an excerpt):
9.4 Solutions of Exercises of This Chapter: Strong Focusing Synchrotron 359

• The transport matrix of BEND is found under MATRIX, bottom of zgoubi.res

.
execution listing:

By comparison, from Eq. 14.7 one gets reasonable agreement, for instance
cos(α − ε) sin(α − ε) + sin ε
.T11 = = 0.9389152, T12 = ρ sin α = 2.42548586, T26 = = 0.3856742
cos ε cos ε

. tan ε
.T33 = 1 − α tan ε = 0.984570, T43 = − (2 − tan ε) = −0.012302
ρ

Note that the wedge angle .ε in .T33 and .T43 , which are the hard edge model coef-
.
ficients, should actually be corrected for the extent of the fringe field (.λ = 8 cm),
Eqs. 14.19, 14.20: the present agreement shows that this effect is marginal.
.• The magnetic field across the bend is produced using the input data file of
Table 9.10, which resorts to OPTIONS[CONSTY ON]. It is displayed in Fig. 9.23.
BEND is defined in a Cartesian frame, its axis is a straight line, thus OPTIONS
[CONSTY], which maintains a constant distance to that axis, results in straight
trajectories. Using DIPOLE instead would result in circular trajectories, at constant
distance from its RM-radius reference arc, as DIPOLE is defined in a polar frame.
.• SATURNE cell is built from the half-arc (Table 9.11) followed by a FODO cell

(Table 9.12). Simulation of a cell follows, Table 9.13, the ring is a 4-cell assembly,
Table 9.14.
The TWISS command performed to determine the optical functions along the
ring (Table 9.14) has the virtue of logging in zgoubi.res execution listing the trans-
port matrices of individual optical elements as the propagation of the optical func-
tions along the sequence proceeds. Looking up zgoubi.res one finds for the focusing
quadrupole for instance:
360 9 Strong Focusing Synchrotron

Table 9.10 Simulation input data file to produce the magnetic field across BEND, on a straight
axis and 5 cm off-mid plane. Another possibility for the same result would be, rather than using
OPTIONS[CONSTY], to launch the two particles with a large relative rigidity D, they would
therefore go straight

Fig. 9.23 Magnetic field

across the bend, in the
mid-plane and 5 cm off. In
the latter case, varying field
causes overshoots in the
fall-off regions. A graph
obtained using zpop: menu
7; 1/1 to open zgoubi.plt;
2/[6, 32] for . B Z versus .s;
7 to plot

Agreement can be observed with transport coefficient values obtained from

Eq. 14.27 taken with . L = 46.723 cm, . K = 0.763695 m−1 , namely,
√ √ √
T = cos L K = 0.9177929, T21 = − K sin L K − 0.3469888
. 11
√ √ √
T33 = cosh L K = 1.084523, T21 = K sinh L K = 0.3668189

Similar agreement is observed for the defocusing quadrupole.

The optical functions of the ring are computed using the input data file of
Table 9.13, they are displayed in Fig. 9.24.
9.4 Solutions of Exercises of This Chapter: Strong Focusing Synchrotron 361

Beam matrix and optical parameters (tunes, momentum compaction, …) are

found in zgoubi.res execution listing resulting from the ring simulation (Table 9.14),

namely:

Table 9.11 Simulation input data file: half SATURNE 2 arc, from center of QF to center of QF; com-
putation of its transport matrix. This input data file saturneHalfArc.inc defines the arc sequence seg-
ment sat_HalfArc_S to sat_HalfArc_E for use in subsequent INCLUDE statements. PUH, HDEF,
HKIC as well as PUV, VDEF, VKIC markers and thin lenses are not used here, they are provisions
for a subsequent SVD orbit correction simulation using SVDOC
362 9 Strong Focusing Synchrotron

Table 9.12 Simulation input data file: SATURNE 2 FODO cell, from center of QF to center of
QF; computation of its transport matrix. This input data file saturneFODO.inc defines the FODO
sequence segment sat_HalfArc_S to sat_HalfArc_E for use in subsequent INCLUDE statements.
PUH, HDEF, HKIC as well as PUV, VDEF, VKIC markers and thin lenses are not used here, they
are provisions for a subsequent SVD orbit correction simulation using SVDOC

(b) Six-D bunch acceleration. Betatron damping.

The input data file for this simulation is given in Table 9.15.
A major change is BORO value, set to 1.03527036749193 T m, i.e. a kinetic energy
of 50 MeV. SCALING is introduced to scale the field in all magnetic elements to that
value of the rigidity. SCALING[NTIM .= −1] will cause fields to further scale to the
increase of the reference rigidity by CAVITE.
Periodic beam parameters under MCOBJET are taken from the beam matrix
resulting from the TWISS run in question (a). CAVITE is programmed with harmonic
◦
.h = 3 and synchronous phase .30 .
Various tracking results are displayed
√ in Fig. 9.25. During acceleration, betatron
oscillation amplitudes damp as .1/ βγ (Eq. 9.40). With all parameters maintained
^ p damps as .1/ βγ 1/4 (Eq. 9.37).
constant except for the energy, ./\p/
9.4 Solutions of Exercises of This Chapter: Strong Focusing Synchrotron 363

Table 9.13 Simulation input data file: SATURNE 2 cell, comprised of two half-arcs and a FODO
cell; computation of its optical functions using TWISS. The gnuplot script gnuplot_TWISS.gnu is
copied from the [pathTo]/branches/zgoubi-code/toolbox/gnuplotFiles/gnuplot_Zfai/ library which
is part of zgoubi sourceforge package

Fig. 9.24 Quasi-zero closed orbit, and the optical functions in SATURNE 2, from a TWISS com-
mand (Table 9.14)

(c) Momentum acceptance. Longitudinal separatrix.

The input data file of Table 9.15 can be used here, mutatis mutandis, as follows.
Longitudinal phase space motion is computed

– over a few thousand turns: use REBELOTE[IPASS.=few thousand],

– for a few particles in the vicinity of the separatrix: use OBJET[KOBJ.=2] to
define individual particles. Particles are launched with null longitudinal offset
(OBJET[. S0 = 0]), dp/p near momentum acceptance. For instance, for the station-
ary bucket case:
364 9 Strong Focusing Synchrotron

Table 9.14 Simulation input data file: SATURNE 2 ring, comprised of 4 cells; computation of its
optical functions using TWISS; a SYSTEM call to a gnuplot script gets them plotted

– the synchronous phase in CAVITE has to be set to proper value.

Tracking hypotheses are recapped in Table 9.16 outcomes are displayed in Fig. 9.16
(Sect. 9.2.4). The paraxial synchrotron tune and bucket height, Table 9.17, can be
checked against Eqs. 9.30, 9.36, respectively.

9.2 Non-Linear Motion in SATURNE 2

(a, b) Third- or fourth-integer phase space portraits.
It is necessary to run the lattice near 3rd integer horizontal tune, .νY ≈ 3 + 23 , taken
near the nominal tune in order to avoid excessive perturbation of the optics. (respec-
tively 4th integer horizontal tune, .νY ≈ 3 + 43 .) The vertical tune is left unchanged.
A preliminary FIT will provide the required SCALING coefficients for
QUADRUPO (the focusing quadrupole family) and MULTIPOL (defocusing
quadrupole family). The file for that is given in Table 9.18. FIT outcomes, found
in the execution listing zgoubi.res, are as follows:
– case of .νY = 3 + 23 :
9.4 Solutions of Exercises of This Chapter: Strong Focusing Synchrotron 365

Table 9.15 Simulation input data file: acceleration from 50 MeV to 3 GeV in SATURNE 2 ring.
To save on CPU-time and on voluminous log file zgoubi.plt, OPTIONS[.plt 0] makes sure
that any . I L /= 0 value in the different magnets (Tables 9.9, 9.11 and 9.12) is overriden with
. I L = 0. This simulation, with 30 particles, takes about 6 min with a 3 GHz clock CPU. The gnu-
plot script gnuplot_Zfai_YZD.vs.turn_multiplot.gnu is copied from the [pathTo]/branches/zgoubi-
code/toolbox/gnuplotFiles/gnuplot_Zfai/ library which is part of zgoubi sourceforge pack-
age Note: this file, and all INCLUDE files it resorts to, are available in zgoubi source-
forge repository at https://sourceforge.net/p/zgoubi/code/HEAD/tree/branches/exemples/book/
zgoubiMaterial/synchrotron_strongFocusing/SaturneII/accelerationCycle
366 9 Strong Focusing Synchrotron

Fig. 9.25 Acceleration over . E : 50 MeV → 3 GeV in SATURNE 2: damped Y, Z motions, and
.D− 1 = δ p/ p, as a function of turn. Turn-by-turn data read from zgoubi.fai

Table 9.16 Stationary bucket tracking: working hypotheses

. E s (MeV) Proton mass + 50
. Bρ (T m) 1.035
h 3
V (kV) 10
.φs (deg) 0, 15, 30 or 60
alpha 0.01505
eta 0.8863
T rev. (.µs) 0.35
RF (MHz) 8.5

Table 9.17 Accelerating bucket: theoretical expectations (left) and tracking outcomes (right). The
numerical value of .νs is obtained as the inverse of the number of turns necessary to close the phase
space ellipse
Theoretical Numerical
Synch. tune .νs (.×10−4 ) (Eq. 9.30) 7.5779 (Eq. 9.30) 7.5758 (Eq. 9.30)
^
/\p −3
.
p (.×10 ) .φs =0 4.9562 5.05

(Eq. 9.36) .φs = 15 3.9249 3.97

.φs = 30 2.9002 2.94
.φs = 60 1.0693 1.10
9.4 Solutions of Exercises of This Chapter: Strong Focusing Synchrotron 367

Table 9.18 Simulation input data file: find focusing and defocusing quadrupole family settings, for
horizontal tune value .νY = 3 + 23 (or .νY = 3 + 43 ). The INCLUDE file saturneCell.inc is defined
in Exercise 9.1-a

– case of .νY = 3 + 43 :

Finally, non-linear horizontal phase space portraits, at fixed energy, are generated
using the input data file given in Table 9.19, essentially a copy of the file in Table 9.15
with the following modifications:

– use OBJET[KOBJ.=2] to define individual particles with diverse invariant values.

The value of the rigidity does not matter;
– modify the SCALING factors for QUADRUPO and MULTIPOL families, so to
set the horizontal tune to a proper value, near 3rd- or 4th-integer;
368 9 Strong Focusing Synchrotron

Table 9.19 Simulation input data file: generate a non-linear phase space, at fixed energy. The input
data file of Table 9.15 is used, with a few modifications. In particular, the QUADRUPO family and
MULTIPOL family SCALING coefficients are updated according to the previous FIT outcomes

– remove CAVITE;
– add a thin lens to excite the resonance. MULTIPOL can be used, with a sextupole
component to excite the 3rd integer resonance motion, near .νY = 3 + 23 (or an
octupole component to excite the 4th integer resonance motion, near .νY = 3 + 43 ).
The resulting phase space portraits are displayed in Fig. 9.26.
9.3 SVD Orbit Correction
An input data file for orbit correction in SATURNE 2, using SVDOC, is given in
Table 9.20. Its data are commented for clarification of the modus operandi. This file
9.4 Solutions of Exercises of This Chapter: Strong Focusing Synchrotron 369

Fig. 9.26 Third-integer and fourth-integer resonance phase space portraits, about.104 turns. Graphs
obtained using zpop: menu 7; 1/5 to open zgoubi.fai; 2/[2,3] for .T versus .Y ; 7 to plot

INCLUDEs SATURNE ring sequence taken from Exercise 9.1, thus it has a 12 PUH
and a 12 PUV family; a 12 HKIC and a 12 VKIC family, thin lenses used to correct
the orbit; a 12 HDEF and a 12 VDEF family, thin lenses used to excite H and V
orbits. The latter are located at respectively focusing and defocusing quadrupoles.
Closed orbit coordinates are found by FIT2, their values at the pick-ups, PUH
and PUV, both before and after correction, are logged in the output file zgoubi.
SVDOrbits.out, which is one of the files resulting from SVDOC execution. This
file also contains the corrector settings. The A matrix is logged in zgoubi.SVD_
Amatrix.out.
The outcomes of two different SVDOC runs, involving 24 PUs and 24 correctors
for one, 24 PUs and 12 correctors for the other, are displayed in Fig. 9.27.

Note: running an SVDOC problem also produces a copy of the original file,
zgoubi.SVD.out.dat, updated with corrector values as found by SVDOC (in a sim-
ilar way that FIT produces a copy of the original problem in zgoubi.FIT.out.dat,
updated with variable values found by FIT). This file zgoubi.SVD.out.dat can be run
as is and will produce the same results as in Fig. 9.27 (make sure it includes a FAI-
STORE[zgoubi.fai PU*] instruction, to save particle data at all PUH and PUV), or
FAISTORE[zgoubi.fai all] instruction, to save particle data at all optical elements).
9.4 Cornell Electron RCS. Radiative Energy Loss
(a) Cornell RCS optics.
The input file in Table 9.8 is run to produce lattice parameters. However, prior to
doing so it is necessary to set the respective quadrupole components of the focusing
and defocusing combined function dipole families (BF and BD) to their expected
nominal value, yielding .νY ≈ 9.62, .ν Z ≈ 13.82 (Table 9.3).
This requires running iteratively, a couple of times,
• the input files in Table 9.4, which computes the orbit across BF and BD so to zero
it in and out,
• the input file in Table 9.8 which, using FIT, computes the BF and BD families field
SCALING coefficients proper to yield the expected tune values.
370 9 Strong Focusing Synchrotron

Table 9.20 Simulation input data file: SVD orbit correction in SATURNE 2 ring, using SVDOC.
saturneCell.inc file is used (Table 9.13), which includes (see Tables 9.11 and 9.12) 12 PUH and 12
PUV; 12 HKIC and 12VKIC (correctors); 12 HDEF and 12 VDEF (excite orbit defects). Comment
for instance every other corrector for a 24 PUs .× 12 correctors system

The reason for this iteration is that, changing the field scaling to match the tunes,
also changes the orbit across the combined function dipoles.
Results are found under TWISS in the execution listing zgoubi.res. They are also
logged in zgoubi.TWISS.out header:
9.4 Solutions of Exercises of This Chapter: Strong Focusing Synchrotron 371

Fig. 9.27 Defect closed orbit observed at PUs, and corrected orbit obtained using SVDOC, both
H and V cases. These data are read from the SVDOC procedure output file zgoubi.SVDOrbits.out.
Left: case of 24 defects (at QF and QD quadrupoles) and 24 correctors at the same locations. Right:
case where every other corrector is removed, leaving only 12 for orbit correction

Fig. 9.28 Quasi-zero closed orbit, and the optical functions in Cornell RCS, from a TWISS com-
mand (Table 9.8)

TWISS logs the optical functions in zgoubi.TWISS.out. A gnuplot script taken

from the sourceforge repository [49] is used to plot them, Fig. 9.28.

(b) An acceleration cycle, without radiative energy loss.

The input data file is given in Table 9.21, it is derived from that in Table 9.8, mutatis
mutandis.
Note in particular “CornellSynch” argument under CAVITE. This causes cavite.f
program to resort to the hard-coded Cornell RCS voltage function [46]
372 9 Strong Focusing Synchrotron

Table 9.21 Simulation input data file for an acceleration cycle in Cornell RCS, from 320 MeV to
∼8 GeV

. V (t) = 4.4 sin(2π f t) + 8.8 sin8 (2π f t/2), f = 60 Hz

SCALING[NT .= −1] ensures that magnetic fields follow the turn by turn rigidity
increase imparted by CAVITE.
The expected particle motion is summarized in Fig. 9.29.

(c, d) Simulation of an acceleration cycle with radiation loss, and spins.

Indications are provided here: detailed simulations are left to the reader. Further
guidance, including methods and data as well as their outcomes, can be found in [46].
The input data file in Table 9.21 is used: comment and uncomment as indicated,
in order to change from ignoring SR to accounting for it. Typical particle dynamics
outcomes expected are displayed in Fig. 9.30.
Non-systematic spin resonances can be excited by introducing a random gra-
dient defect in the combined function dipoles. This is achieved by adding, before
INCLUDE, an ERRORS command, as follows:
9.4 Solutions of Exercises of This Chapter: Strong Focusing Synchrotron 373

Fig. 9.29 Transverse and longitudinal phase spaces, 2,300 turns from 0.32 to 7.2 GeV, SR ignored.
Motion damping is apparent in the three phase spaces

Fig. 9.30 Particle motion during acceleration in Cornell RCS, from injection energy, 0.32 GeV, to
up to 8 GeV [46]. The radial motion features anti-damping

The amplitude of the random defect,.10−2 relative, here, can be changed to control
the strength of the resonances.
Tracking is expected to show drops of average polarization, at .aγ locations which
coincide with the resonance line spectrum. The latter can be computed using Eq. 9.53
(with .a ≡ G = 1.159652 × 10−3 ). The optical functions and quadrupole strengths
needed for that are read from the zgoubi.TWISS.out file produced in question (a). In
the present simplified lattice, the first strong depolarizing resonance is at .aγ = ν Z =
13.82. Figure 9.31 displays a simulation of resonance crossing during acceleration,
typical of expected outcomes.

9.5 Coupling in a Light Source Storage Ring

A JINST article, “Simulation of radiation damping in rings, using stepwise ray-
tracing methods” [48], provides all necessary details regarding the working hypothe-
ses, and guidance regarding damping simulations, based on the early ESRF lattice.
374 9 Strong Focusing Synchrotron

Fig. 9.31 Depolarizing

resonance crossing in
Cornell RCS, case of a few
electrons, launched on
invariant .βγε y /π =
100 µm [46]. The lattice
includes errors: a 1.2 mm
rms vertical closed orbit, a
gradient defect .d K 1/K 1 in
.±1% and a . K 0 roll in
◦
.±0.02 , both random
uniform, in all main bends

Regarding the present case of a coupled lattice, Sect. 5 in that article is concerned.
Under graphical form, Figs. 7 and 8 should be reproduced. If so, then raytracing
outcomes are in accord with Eqs. 24–26.
In addition to referring to Zgoubi Users’ Guide [50], general guidance regarding
the present simulations, the way to handle zgoubi i/o files for instance, the various
options, including SRLOSS, can be found in the previous exercises, and in simulation
examples found in zgoubi sourceforge repository, in the folder https://sourceforge.
net/p/zgoubi/code/HEAD/tree/branches/exemples/.
Working the details of the present simulations is left to the reader.

9.6 SR Electric Impulse and Interference in a Miniwiggler

A program can be written for the calculation of the electric field impulse and its
spectrum, based on the equations given in Sect. 8.2.3. Particle data can be read from
zgoubi.plt. On the other hand, zpop does perform SR computations (its menu 8/16), it
is used here. Zpop’s file sref.f manages the computation of the electric field impulse,
using Eq 8.42, accounting for Eq. 8.37 for the retarded time; the file srdw.f manages
the computation of the spectral brightness, using Eqs. 8.38 and 8.39.

(a) Simulation of a pair of dipoles, configured as in the LEP miniwiggler.

The input data file is given in Table 9.22. The trajectory so obtained is displayed in
Fig. 9.32.

(a) Single dipole. Electric field impulse, spectral brightness.

Using Table 9.22 data file, a way to account for radiation from a single dipole is to set
MULTIPOL[IL.=2] in one or the other. Or insert END (or FIN, in French), following
the first dipole.
The electric field impulse . E σ (φ = ψ = 0, t) from the first dipole, in observer
time, is displayed in Fig. 9.33. It would have opposite sign from the second dipole.
. E π (t) = 0 in the bend plane.
9.4 Solutions of Exercises of This Chapter: Strong Focusing Synchrotron 375

Table 9.22 Simulation input data file for a pair of dipoles. Initial electron incidence is .α/2, so that
the observation direction .φ = ψ = 0 is tangent to the trajectory arc in the center of a dipole

Fig. 9.32 Electron

trajectory through a pair of
dipoles of opposite signs. A
graph obtained using zpop:
menu 7; 1/1 to open
zgoubi.plt; 2/[6,2] for .Y
versus .s; 7 to plot

Expected duration of the impulse: the typical width of the impulse in observer
time, in the bend plane (.ψ = 0) is (Eq. 8.44)./\τc = 2ρ/(3γ 3 c) ≈ (2 × 657)/(3(2000
E [GeV ] )3 c) ≈ 3 × 10−21 s. This is consistent with . E σ (τ ) duration in Fig. 9.33. The
spectrum in the bend plane is expected to peak at (Eq. 8.44) .hωc = h/(2π/\τc ) ≈
0.3[s] h/e ≈ 0.2 MeV. This is consistent with the spectrum result in Fig. 9.33.

(b) Interference between two dipoles. Electric field impulse, spectral brightness.
Case of bend plane radiation: the results are displayed in Fig. 9.34.
376 9 Strong Focusing Synchrotron

Fig. 9.33 Single dipole electric impulse. E σ (τ ) observed in the direction.φ = 0, ψ = 0, and its spec-
trum. . E π (τ ) ≡ 0 in the bend plane. Graphs obtained using zpop: menu 8; 16 will open zgoubi.plt;
2/[1,E] for particle 1, electron; 5 for electric impulse or 6 for spectrum

Fig. 9.34 Dipole doublet electric impulse observed in the direction .φ = 0, ψ = 0 (left), and its
interferencial spectrum - the single dipole spectrum is superimposed for comparison (right). Graphs
obtained using zpop: menu 8; 16 will open zgoubi.plt; 2/[1,E] for particle 1, electron; 5 for electric
impulse or 6 for spectrum

Case of radiation observed in the direction .φ = 0, ψ = 0.2 mrad: the results are
displayed in Fig. 9.35.

(b) Interference. Same sign dipoles.

The input data file of Table 9.22 is used, changing the magnetic field sign in the first
dipole, to negative, and the launch angle of the particle under OBJET, to .−0.8 mrad.
The resulting trajectory is displayed in Fig. 9.36.
The resulting electric impulses and spectra in the direction .φ = 0, ψ = 0.2 mrad
are displayed in Fig. 9.37.
9.7 Depolarizing Resonances in SATURNE 2
Crossing .Gγ = 7 − ν Z non-systematic spin resonance
The input data file is copied from Table 9.15, with the following modifications:
– use OBJET[KOBJ.=2] to define individual particles. It does not matter if initial
coordinates are not taken on respective chromatic closed orbits, as the resonance
strength does not depend on the horizontal invariant value;
9.4 Solutions of Exercises of This Chapter: Strong Focusing Synchrotron 377

Fig. 9.35 Left: electric impulse components . E σ (τ ) and . E π (τ ) from a dipole doublet in observer
time, in the direction .φ = 0, ψ = 0.2 mrad. Right: their interferencial spectra

Fig. 9.36 Electron trajectory through a pair of same sign dipoles. A graph obtained using zpop:
menu 7; 1/1 to open zgoubi.plt; .2/[6, 2] for .Y versus .s; 7 to plot

Fig. 9.37 Left: electric impulse components. E σ (τ ) and. E π (τ ) from a doublet of same sign dipoles,
in observer time, in the direction .φ = 0, ψ = 0.2 mrad. Right: their interferencial spectra. This is
an instance of edge radiation, interference is between impulse from downstream end of upstream
dipole and from upstream end of downstream dipole
378 9 Strong Focusing Synchrotron

Fig. 9.38 Acceleration through .Gγ = 7 − ν Z spin resonance: Y, Z particle motion, and . D − 1 =
δ p/ p, as a function of turn. Turn-by-turn data read from zgoubi.fai

– change the rigidity to 5.01 T m, and accordingly the SCALING factors to the same
value; this sets the starting point of the tracking a few 1,000 turns upstream of the
resonance, where its effect is not felt;
– add SPNTRK[KSO.=3] to set initial spin of all particles to vertical;
– add a thin lens to break the 4-periodicity—necessary condition to obtain any depo-
larization at traversal of .Gγ = 7 − ν Z , as it is a non-systematic spin resonance.

The resulting resonance crossing input data file is given in Table 9.23.
Tracking results are summarized in Figs. 9.38 and 9.39. Spin motion features
multiple resonance crossing in the three cases .dp/ p /= 0 (Fig. 9.40), an effect of the
synchrotron motion.
9.4 Solutions of Exercises of This Chapter: Strong Focusing Synchrotron 379

Table 9.23 Simulation input data file: acceleration through.Gγ = 7 − ν Z spin resonance. To avoid
any time-consuming logging to zgoubi.plt, OPTIONS[.plt=0] imposes . I L = 0 in all magnets
380 9 Strong Focusing Synchrotron

Fig. 9.39 Spin motion

through .Gγ = 7 − ν Z spin
resonance. A graph obtained
using zpop: menu 7; 1/5 to
open zgoubi.fai; 2/[39,23]
for . S Z versus turn; 7 to plot

Fig. 9.40 Multiple

resonance crossing due to
energy oscillation. The spin
resonance is at
.Gγ = 7 − ν Z ≈ 3.39, i.e.
. E ≈ 836 MeV (horizontal
line)

References

1. N. Christofilos, Focussing system for ions and electrons. US Patent Office Application filed
March 10, 1950, Serial No. 148,920. https://patentimages.storage.googleapis.com/fa/bb/52/
0ce28e28b492a6/US2736799.pdf
2. Credit: Brookhaven National Laboratory. https://www.flickr.com/photos/brookhavenlab/
8495311598/in/album-72157611796003039/
3. E.D. Courant, M.S. Livingston, H.S. Snyder, The strong-focusing synchrotron - a new high
energy accelerator. Phys. Rev. 88, 1190 (1952)
4. E.D. Courant, H.S. Snyder, Theory of the alternating-gradient synchrotron. Ann. Phys. 3, 1–48
(1958)
5. SATURNE 2 photo: credit CEA Saclay. Archives historiques du CEA. Copyright CEA/Service
de documentation
6. Credit: Lawrence Berkeley National Laboratory. The Regents of the University of California,
Lawrence Berkeley National Laboratory
7. Radial focusing in the linear accelerator. Phys. Rev. 88(5) (1952)
8. M. Benedikt, F. Zimmermann, Status of the future circular collider study, in TUYMH01
Proceedings of RuPAC2016, St. Petersburg, Russia. https://accelconf.web.cern.ch/rupac2016/
papers/tuymh01.pdf
References 381

9. F. Méot, et al., Progress on the optics modeling of BMI’s ion rapid-cycling medical synchrotron
at BNL, in THPMP050, 10th International Particle Accelerator Conference IPAC2019, Mel-
bourne, Australia. https://accelconf.web.cern.ch/ipac2019/papers/thpmp050.pdf Copyrights
under license CC-BY-3.0, https://creativecommons.org/licenses/by/3.0; no change to the mate-
rial
10. N. Nishimori, A new compact 3 GeV light source in Japan, in 13th International Particle Accel-
erator Conference IPAC2022, Bangkok, Thailand. https://accelconf.web.cern.ch/ipac2022/
papers/thixsp1.pdf
11. F. Méot, eRHIC ERL modeling in Zgoubi. BNL-111832-2016-TECH; EIC/49;BNL-111832-
2016-IR. https://technotes.bnl.gov/PDF?publicationId=38865. https://www.osti.gov/biblio/
1335396
12. C. Benabderrahmane, Status of the ESRF-EBS magnets, in WEPMK009, 9th International
Particle Accelerator Conference, IPAC2018, Vancouver, BC, Canada. https://accelconf.web.
cern.ch/ipac2018/papers/wepmk009.pdf
13. F. Méot, et al.: Plans for polarized bunch R&D at Cornell rapid-cycling synchrotron,
in MOPMF013, 9th International Particle Accelerator Conference IPAC2018, Vancouver,
BC, Canada. https://accelconf.web.cern.ch/ipac2018/papers/mopmf013.pdf Figure 9.7: Copy-
rights under license CC-BY-3.0, https://creativecommons.org/licenses/by/3.0; no change to the
material
14. R.R. Wilson, The 10 to 20 GeV Cornell electron synchrotron. Report CS-33, Laboratory of
Nuclear Studies, Cornell University, Ithaca, New York (May 1, 1967)
15. D.L. Rubin, et al., Upgrade of the Cornell electron storage ring as a synchrotron light source,
in WEPOB36, Proceedings of NAPAC2016, Chicago, IL, USA. https://accelconf.web.cern.ch/
napac2016/papers/wepob36.pdf
16. CERN Courrier: Partnership yields big wins for the EIC. 27 September 2021 issue. https://
cerncourier.com/a/partnership-yields-big-wins-for-the-eic/
17. https://www.bnl.gov/eic/
18. https://www.slac.stanford.edu/gen/grad/GradHandbook/slac.html
19. https://www.desy.de/~teslatdr/tdr_web/pages/latest_version.html
20. https://www-project.slac.stanford.edu/lc/
21. https://muoncollider.web.cern.ch/node/25
22. https://linearcollider.org/
23. N.V. Litvinenko, N. Bachhawat, M. Chamizo-Llatas, Y. Jing, F. Méot, I. Petrushina, T. Roser,
The ReLiC: recycling linear e+e- Collider. arXiv:2203.06476 [hep-ex]
24. G. Jackson (Ed.), Fermilab recycler ring technical design report. Rev. 1.1. FERMILAB-TM-
1981 (July 1996). http://inspirehep.net/record/424541/files/fermilab-tm-1981.PDF
25. F. Méot, On the Effects of Fringe Fields in the Recycler Ring. FERMILAB-TM-2016
(Aug. 1997). http://inspirehep.net/record/448603/files/fermilab-tm-2016.PDF
26. G. Leleux, Compléments sur la Physique des Accélérateurs. DEA “Physique et Technologie des
Grands Instruments”, Université Paris VI. Rapport interne LNS//86-101, CEA Saclay (1986)
27. F. Méot, et al., Beam dynamics validation of the Halbach technology FFAG cell for Cornell-
BNL energy recovery Linac. Nuclear Inst. Methods Phys. Res. A 896, 60–67 (2018)
28. H. Bruck, Accélérateurs circulaires de particules. Presses Universitaires de France (1966)
29. G. Leleux, Accélérateurs Circulaires. INSTN lectures, internal report CEA Saclay (1978),
unpublished
30. B.J. King, Further studies on the prospects for many-TeV muon colliders, in Proceedings
of PAC 2001 Particle Accelerator Conference, 18–22 June 2001, Chicago, IL, USA. https://
accelconf.web.cern.ch/p01/PAPERS/RPPH314.PDF
31. A. Hofmann, The Physics of Synchrotron Radiation. Cambridge Monographs on Particle
Physics, Nuclear Physics and Cosmology (20) (Cambridge University Press, 2004)
32. G. Leleux, Rayonnement synchrotron (Aspect machine). Note technique, Laboratoire National
SATURNE, CEA Saclay (1993) (unpublished)
33. A. Hofmann, F. Méot, Optical resolution of beam cross-section measurements by means of
synchrotron radiation. Nucl. Inst. Methods 203, 483–493 (1982)
382 9 Strong Focusing Synchrotron

34. R. Bossart, et al., Proton beam profile measurements with synchrotron light. CERN-SPS-80-
8-ABM, 18 June 1980
35. F. Méot, Mesure de profil par rayonnement ondulateur des faisceaux de protons et antiprotons.
Ph.D. Thesis. Report CERN/SPS 81-21 (ABM) 30 October 1981
36. L. Ponce, R. Jung, F. Méot, LHC proton beam diagnostics using synchrotron radiation. Yellow
Report CERN-2004-007
37. F. Méot, L. Ponce, N. Ponthieu, Low frequency interference between short SR sources. PRST-
AB 4, 062801 (2001)
38. F. Méot, A theory of low frequency far-field synchrotron radiation. Part Accel 62, 215–239
(1999)
39. F. Méot, Synchrotron radiation interferences at the LEP miniwiggler. CERN SL/94-22 (AP)
(1994)
40. F. Méot, Polarized Beam Dynamics and Instrumentation in Particle Accelerators. USPAS
Summer 2021 Spin Class Lectures, Open Access (Springer Nature, 2023). https://link.springer.
com/book/10.1007/978-3-031-16715-7
41. S.Y. Lee, Spin Dynamics and Snakes in Synchrotrons (World Scientific, 1997)
42. The 20 years of the synchrotron SATURNE-2, in Proceedings of the Colloquium, Paris, France,
04–05 May 1998, ed. by A. Boudard, P.-A. Chamouard. Edited By CEA - Laboratoire National
SATURNE & CEN Saclay, France. https://doi.org/10.1142/3965
43. Plus d’anneaux autour de SATURNE (pp. 33–34) Published in: Courrier CERN Volume 39,
Num. 2, Mars 1999. https://cds.cern.ch/record/1740121
44. E. Grorud, J.L. Laclare, G. Leleux, Crossing of depolarization resonances in strongly modulated
structures. IEEE Trans. Nucl. Sci. NS-26(3) (1979). https://accelconf.web.cern.ch/p79/PDF/
PAC1979_3209.PDF
45. J.P. Aknin, et al., Status report on rejuvenating SATURNE and future aspects. PAC 1979
Conference. IEEE Tans. Nucl. Sci. NS-26(3) (1979). https://accelconf.web.cern.ch/p79/PDF/
PAC1979_3138.PDF
46. F. Méot, Polarized e-bunch acceleration at Cornell RCS: tentative tracking simulations. Tech.
Note BNL-114452-2017-TECHEIC/57;BNL-114452-2017-IR (2017). https://technotes.bnl.
gov/PDF?publicationId=42654https://www.osti.gov/biblio/1408712
47. USPAS Summer 2021 Spin Class Lectures, Polarized Beam Dynamics and Instrumentation
in Particle Accelerators, ed. by F. Méot et al. (Springer, Particle Acceleration and Detection,
2022). https://doi.org/10.1007/978-3-031-16715-7
48. F. Méot, Simulation of radiation damping in rings, using stepwise ray-tracing methods. 2015
JINST 10 T06006. http://iopscience.iop.org/1748-0221/10/06/T06006
49. Gnuplot scripts to plot optical functions, reading from zgoubi.TWISS.out, can
be found at https://sourceforge.net/p/zgoubi/code/HEAD/tree/trunk/toolbox/gnuplotFiles/
gnuplot_TWISS/
50. F. Méot, Zgoubi users’ guide. https://www.osti.gov/biblio/1062013-zgoubi-users-guide.
Sourceforge latest version: https://sourceforge.net/p/zgoubi/code/HEAD/tree/trunk/guide/
Zgoubi.pdf. The betaFromPlt.f program is available here: https://sourceforge.net/p/zgoubi/
code/HEAD/tree/trunk/toolbox/betaFromPlt/
References 383

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0
International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing,
adaptation, distribution and reproduction in any medium or format, as long as you give appropriate
credit to the original author(s) and the source, provide a link to the Creative Commons license and
indicate if changes were made.
The images or other third party material in this chapter are included in the chapter’s Creative
Commons license, unless indicated otherwise in a credit line to the material. If material is not
included in the chapter’s Creative Commons license and your intended use is not permitted by
statutory regulation or exceeds the permitted use, you will need to obtain permission directly from
the copyright holder.
Chapter 10
FFAG, Scaling

Abstract This chapter is an introduction to Fixed-Field Alternating Gradient (FFAG)

cyclic accelerators. It begins with a brief reminder of the historical and technolog-
ical context, and continues with the theoretical material needed for the simulation
exercises. It relies on charged particle optics and acceleration concepts introduced
in the previous cyclotron and synchrotron chapters. Furthermore it addresses

– design aspects of scaling FFAGs,

– beam dynamics in radial- and spiral-sector rings,
– synchrotron acceleration and various other acceleration techniques.
Simulations introduce dedicated keywords providing an analytical modeling of the
field: FFAG (radial sector dipole) and FFAG-SPI (spiral sector). They otherwise use
optical elements met in the previous chapters: DIPOLE[S], TOSCA, CAVITE, data
input/output keywords such as FAISCEAU, FAISTORE, the SYSTEM keyword, etc.
Beam dynamics simulations include
– particle trajectories through multiple-dipole FFAG cells,
– closed-orbit finding, from multi-turn raytracing or using FIT,
– deriving ancillary outcomes from rays, such as
• transport matrices using MATRIX,
• periodic optical functions and their transport using TWISS,
– finding dynamical aperture.

Notations Used in the Text

. A Sector angle of a dipole

. _ . Bx,y,s ; . B0
B; Field vector; components in moving frame; field at reference radius
. R0
Bρ; Bρ0
. Particle rigidity: . Bρ = p/q;
{ for reference momentum . p0
C; .C0
. Closed orbit length:.C = ds = 2π R; for reference momentum. p0
© The Author(s) 2024 385
F. Méot, Understanding the Physics of Particle Accelerators, Particle Acceleration
and Detection, https://doi.org/10.1007/978-3-031-59979-8_10
386 10 FFAG, Scaling

.ds Path length increment: .ds = Rdθ

. E; . E xtr ; . E in j ; . E s Particle energy, . E = γ m 0 c2 ; at extraction; injection;
synchronous
EFB Effective field boundary
. f, F (θ ), F (r, θ ) Flutter factor
. f rev , . f rf Revolution and RF voltage frequencies
. g; g0 Gap of a dipole magnet, a function of R; .g(R = R0 )
.h Harmonic number, an integer, .h = f rf / f rev
. I1 Fringe field integral
.k Geometric field index: .k = BR ∂∂ BR ≈ −n ρR
.L Magnetic length
.m; .m 0 ; M Particle mass; rest mass; mass in .eV /c2 units
.N Number of cells in a ring
.n Local focusing index: .n = − ρB ∂∂ Bx ≈= − ρR k
.pf Packing factor: .pf = L/C
. p; p0 ; δp, /\p Particle momentum; reference momentum; offset
.q Particle charge
.R Average closed orbit radius: . R = C / 2π
. R; R0 Radial coordinate, from center of ring; reference
.R F Radio-Frequency
.s Path variable
.v Particle velocity
. Vrf ; . V̂rf Accelerating voltage; peak value
, ,
[ ]
.x, x , y, y Particle coordinates in the moving frame . (∗), = d(∗)/ds
.α Momentum compaction, or trajectory deviation
.β = v/c; .β0 ; .βs Normalized velocity; reference; synchronous
.βu , αu , γu ; .ηu Optical functions (.u : x, y, l); dispersion function
.γ Lorentz relativistic factor: .γ = E/m 0 c2 = E[eV ]/M
.δ Relative momentum offset: .δ = δp/ p
.ε Wedge angle
.E R Strength of a depolarizing resonance
.εu Courant-Snyder invariant; beam emittance{(.u : x, y, l)
.μu Betatron phase advance per period, .μu = period βuds(s) (.u : x, y)
.νu Wave numbers, horizontal, vertical, synchrotron (.u : x, y, l)
.ζ Spiral angle of a spiral sector dipole EFB
.η Phase-slip factor
.θ Azimuthal angle
.κ Gap shape index
.λ Fringe field extent
.ωrf Accelerating voltage angular frequency: .ωrf = 2π f rf
.ρ Local curvature radius
.ϕ Scalloping angle
.φs Synchronous RF phase
10.1 Introduction 387

10.1 Introduction

The Fixed field alternating gradient (FFAG) concept was devised in the early
1950s [1–6]. Electrostatic accelerators, cyclotrons, betatrons, synchrotrons were part
of the landscape at the time, as instruments for nuclear physics research, medical and
industrial applications, X-ray generators, etc. Higher energies were driving acceler-
ator technology R&D, and strong focusing, pulsed synchrotron cascades and col-
lider rings were on their way to take over. The FFAG concept was explored as an
alternate implementation of strong focusing, liable to allow high intensity beams
due to their—synchrocyclotron-like—capability of very fast cycling resulting from
the fixed magnetic field, and to the large momentum and geometrical acceptance of
strong focusing zero-chromaticity optics. Three electron models were built and oper-
ated in the 1953–1967 period (Fig. 10.1), by the Midwestern Universities Research
Association [1]. These early FFAG studies produced a wealth of theoretical and com-
putational contributions to beam theory and beam manipulation in cyclic accelerator
magnets and RF systems.
The interest in FFAG technologies resurrected in the late 1990s in the context
of high energy physics R&D programs and the acceleration of short-lived beams,
with potential spin-offs such as medical and other high power proton and elec-
tron beam accelerators [7–9] (Fig. 10.2). Several proton and electron machines were
built in Japan from the 1990s on [7]. These developments included an ADS-Reactor

Fig. 10.1 The third FFAG electron model at MURA, a two-way 50 MeV electron ring [5]. The ring
was operated in a collider mode with two counter-rotating beams, in the early 1960s
388 10 FFAG, Scaling

Fig. 10.2 PoP, the first proton FFAG, a 500 keV Proof-of-Principle ring operated at KEK from 1999
on [10]. Beam from the 50 kV . H + source (in the background) is steered across the ring vacuum
chamber onto the inner radius injection orbit. The high gradient RF cavity can be seen between
dipoles to the right, with its power supply to its right

installation where first experiments in the world for basic ADS research have been
undertaken in 2009, an internal target experiment, and more [11, 12]. A prototype
FFAG spiral sector dipole was built in 2005, as part of a multiple-beam protontherapy
ring design study [13] (Fig. 10.3).
During acceleration orbit spirals out in an FFAG, as in cyclotrons and synchro-
cyclotrons. FFAGs optics is non-isochronous: the radial field index .k is constant to
ensure constant tunes as the beam spirals out (isochronism requires .k ∝ γ 2 , Eq. 4.1).
FFAGs are normally operated as synchrocyclotrons: the RF is cycled, voltage fre-
quency is modulated during the ramp; repetition rates of tens of kHz are potentially
achievable with today’s RF system technologies. High power and fast acceleration
R&D have produced alternate acceleration and RF manipulation techniques, includ-
ing quasi-isochronous optics and CW acceleration [14–18].
10.2 Basic Concepts and Formulæ 389

Fig. 10.3 Full scale prototype (left), and pole (right), of the spiral sector dipole of a proton FFAG
designed for proton-therapy use [13]

10.2 Basic Concepts and Formulæ

Consider Hill’s equations

⎧ 2
⎪
⎪ d x R2
⎪
⎨ dθ 2 + (1 − n)x = 0
ρ2
. (10.1)
⎪
⎪ d2 y R2
⎪
⎩ + 2 ny = 0
dθ 2 ρ

These equations assume that orbit scalloping around the ring is marginal, i.e. a quasi-
circular closed orbit [19]. At a given angle .θ in an FFAG sector, constant radial and
axial focusing is equivalent to
( )| ( 2)|
d R 2 || d R |
. (1 − n) 2 | = 0 and n 2 || = 0 (10.2)
dp ρ θ dp ρ θ

A sufficient condition for Eq. 10.2 is

( )| |
∂ R || ∂n ||
=0 and =0 (10.3)
ρ |θ ∂ p |θ
.
∂p

The first condition yields constant ratio particle radial position/local curvature radius:
the geometrical scaling property, orbits of different momenta scale with energy, the
390 10 FFAG, Scaling

center of similitude is the center of the ring. The second condition yields momentum-
independent local focusing index: the zero-chromaticity property.
Geometrical similarity results in a constant geometrical field index

R ∂ B(R) || R
.k= | ≈ − n = constant (10.4)
B(R) ∂ R R ρ

R
Note that the approximation .k ≈ − n results from the hypothesis of a circular orbit,
ρ
neglecting the orbit scalloping [19, Eq. 8]. Integration yields the R-dependence of
the field in an FFAG dipole,
( )k
R
. B(R) = B0 , k>0 (10.5)
R0

From .k = constant (Eq. 10.4) it results that reversing the sign of the curvature radius
ρ reverses the sign of the field index.n. Radial FFAG lattices combine such alternating
.
index dipoles, Fig. 10.4 and Sect. 10.2.1. A way to obtain such radial field distribution
is by shaping the dipole gap, according to
( )κ
R0
. g(R) ≈ g0 with κ ≈ k (10.6)
R

The gap is greater (lower) at lower (greater) energy and radius (Fig. 10.4). Another
way is by distributed current windings along the poles of a parallel gap dipole [1,
20, 21]. More generally, in a lattice comprised of bends and field-free sections, the
magnetic field along an orbit in the median plane (. y = 0) satisfies
( )k
R
. B(R, θ ) = B0 F (R, θ ) (10.7)
R0

where the .2π/N -periodic flutter factor .F (R, θ ) describes the azimuthal modulation
of the field along an orbit (in a similar way to the modeling of the AVF cyclotron, cf.
Chap. 4, Eqs. 4.4, 4.5 and 4.11).
Orbits
It results from the scaling field (Eq. 10.5) that the average orbit radius and the orbit
length satisfy the momentum dependence, respectively,
( )1/(k+1) ( )1/(k+1) ( ) k+1
1
R( p) Bρ p p
. = = and C( p) = C0 (10.8)
R0 Bρ0 p0 p0

The . R- and .ρ-radius arcs share a common cord (Fig. 10.5), which writes

. R sin(A/2) = ρ sin(π/N ) (10.9)

10.2 Basic Concepts and Formulæ 391

B Positive bend, Negative bend,

converging sector, diverging sector,
F vertically diverging. vertically converging.

I I
F
Toward Small gap, Toward Small gap,
center strong field, center strong field,
of ring Large gap, high energy. of ring Large gap, high energy.
low field, B low field,
low energy. low energy.

Fig. 10.4 Focusing and defocusing FFAG sectors

Fig. 10.5 Geometrical

parameters in a radial
(straight EFBs) or spiral
(dashed lines; spiral angle .ζ )
FFAG sector dipole. O is the
center of the ring and the
EFBs form a sector angle . A.
The . R-radius arc is a line of
constant field (Eq. 10.5). The
closed orbit is approximately
along the .ρ-radius arc. Both
arcs share the same cord

A packing factor can be defined,

magnetic length L N×A

. pf = = = (10.10)
orbit length C 2π

In a general manner, closed orbits have to be computed numerically, searching for

the momentum-dependent closed solution over a period. They feature small ampli-
tude scalloping in the vicinity of an average circular path with radius. R( p) (Eq. 10.8),
thus the initial radius value for a numerical search can be taken as. R ≈ R( p), whereas
the incidence .d R/ds is null with proper choice of the origin.
The orbit excursion, from injection to extraction momentum, satisfies (Eq. 10.8)
( ( 1 )
) 1+k
pin j
. Rxtr − Rin j = Rxtr 1 − (10.11)
pxtr
392 10 FFAG, Scaling

Focusing
There are two ways that the FFAG technique implements strong focusing,

• one consists in alternating strong transverse gradients (large .|n|, Eq. 10.4), which
is achieved by alternating positive- and negative-bend magnets (Sect. 10.2.1), with
the detrimental effect of increased circumference of the ring and decreased packing
factor (Eq. 10.10);
• a second method consists in using positive bend only, and relying on spiral EFBs
and Thomas (AVF) focusing: a large spiral angle (strong axial focusing, radially
defocusing) compensates the large field index (strong radial focusing, axially defo-
cusing). A logarithmic spiral edge [3] has the virtue of ensuring constant wedge
angle (cf. Sect. 10.2.2).

Fringe fields may have a noticeable effect on the effective axial focusing (cf.
Sect. 14.4.1): Eq. 14.20 indicates that, in the case of constant wedge angle (spiral
scaling dipoles), the axial focusing correction for the fringe field extent,.ψ, is constant
iff .λ ∝ R, which requires .λ (a measure of the the gap height) to increase linearly
with radius. In the gap shaping method (Eq. 10.6) the gap decreases with radius
instead, thus leading to an increase in axial wave number with energy: overcoming
that effect requires proper counter-measures such as for instance a specific design of
the chamfers, and field clamps [22].
Wave Numbers
The wave numbers of a radial sector lattice can be derived from the method of aver-
ages, and are given to reasonable accuracy, accounting for orbit scalloping, by [19]
[ ( )2 { } { }2 ]1/2
ν R = 1 + k + k + 23 φ 2 + 98 N 2 φ 2
. [ { } ( )2 { }]1/2 (10.12)
ν y = −k + φ ,2 + k + 21 φ 2

where . N is the number of periods, .φ(θ ) is the scalloping angle (the local angle, at
azimuth .s = Rθ , between the radial axis of the moving frame and the radius to the
center of the ring [19, Fig. 2]), .φ , = dφ/dθ , and .< ∗ > denotes the average value
over a closed orbit.
Keeping the first oder terms only, and accounting for the spiral focusing, leads to
the approximations for the cyclotron, namely (Sects. 4.2.1 and 4.2.2)
√ /
ν ≈
. R 1 + k, νy ≈ −k + F 2 (1 + 2 tan2 ζ ) (10.13)

(with spiral angle .ζ = 0 in the case of a radial sector). The later underestimate the
value of .ν R and overestimate the value of .ν y . However they may be found helpful in
evaluating the relative effects of a small change in the flutter . F, in the geometrical
field index .k, or in the spiral angle .ζ in the case of a spiral sector.
10.2 Basic Concepts and Formulæ 393

10.2.1 Radial Sector

A radial sector scaling FFAG facility is displayed in Fig. 10.6 [23]: a 150 MeV ring
built and operated at KEK in the early 2000s. The ring is comprised of 12 defocusing-
focusing-defocusing (DFD) dipole triplets. The radial dependence of the magnetic
field in the D and F sectors satisfies the scaling law, Eq. 10.5, as a result of the gap
shape, Eq. 10.6. The main parameters of the ring are summarized in Table 10.1.
Hall-probe measurements of an isolated dipole triplet are displayed in Fig. 10.7.
Mutual influence in the ring actually produces a 200 Gauss field across the drift
between two triplets. Figure 10.7 shows the field from OPERA computation in the
periodic hypothesis [24].
Transverse Acceptance
Scaling FFAG optics features large dynamical transverse acceptance [25], see the
case of KEK FFAG ring in Fig. 10.8.

10.2.2 Spiral Sector

The spiral sector FFAG was devised by the MURA group, and an electron model
was operated in 1957 (Fig. 10.9). Compared to the radial sector it had an advantage
of compactness, as focusing relies on the sector edges rather than on alternating
gradient and curvature.
A typical design of a proton spiral sector scaling FFAG is shown in Fig. 10.10: a
variable energy and multiple extraction ring, aimed at cancer tumor treatment [26,
27]. Table 10.2 summarizes the parameters of the dipole magnet and of the ring. The
ring is comprised of 10 spiral sector cells. The radial dependence of the magnetic field
in the spiral dipole satisfies Eq. 10.5 and results from the gap shape which follows
Eq. 10.6. The dipole can be operated up to 2 T on the extraction radius, corresponding

Fig. 10.6 Left: KEK 150 MeV 12-cell scaling FFAG ring, and its cyclotron injector [10]. Right:
Its lattice cell magnet, a DFD dipole triplet. The gap shape follows Eq. 10.6 so ensuring the scaling
field law (Eq. 10.5) [23]
394 10 FFAG, Scaling

Table 10.1 Parameters of KEK 150 MeV radial sector FFAG [23]
Injection—extraction energy MeV 12–150
(proton)
Injection—extraction radius m 4.7–5.2
Lattice DFD
Number of cells . N 12
Maximum .β R ; βz max. m 3.8; 1.3
Wave numbers, .ν R ; νz 3.7; 1.2
Magnet
Type Radial sector DFD triplet
Sector angle . A D ; A F deg 3.43; 10.24
Injection—extraction gap cm 20–4
height
Scaling index .k D = k F 7.6
. B D ; B F , on 150 MeV orbit T .−1.21745; 1.69056

Acceleration
Frequency swing MHz 1.5–4.6
Harmonic 1
Voltage, peak-to-peak kV 19
Cycle time ms 4
Maximum repetition rate Hz 250
Equivalent dB/dt T/s 280
Synchrotron tune .νs 0.039–0.012

Fig. 10.7 Left: measured vertical magnetic field . B Z (X ), along .R = constant arcs across one half
of the radial sector dipole triplet [23]. The X coordinate is along an axis normal to the vertical
symmetry plane of the triplet (the .X = 0 plane). The field on the plateau accurately follows the . R k
scaling law (Eq. 10.5), lower (greater) field at lower (greater) energy, greater (lower) dipole gap.
Right: field along the periodic orbits across the cell at various energies (proton), from an OPERA
field map of the KEK FFAG dipole triplet cell [24]
10.2 Basic Concepts and Formulæ 395

Fig. 10.8 Radial motion 1,000-turn stability limit at various energies. The small ellipse within the
10 MeV stability invariant on the left is for a nominal .ε R = 0.04 π cm beam at injection energy.
The single-cell radial wave number at stability limit (.ν R ) and paraxial (.νr , between parentheses)
give an idea of the amplitude detuning [24]

Fig. 10.9 The second (of three) electron model of an FFAG, a spiral sector design by the MURA
group, operated in 1957 [5]
396 10 FFAG, Scaling

Fig. 10.10 Left: RACCAM proton therapy scaling FFAG ring design, including a variable energy
H.− cyclotron injector. Right: a scheme of its spiral dipole half-yoke, showing the gap shaping
pole piece with its variable width chamfers and the EFB field clamps, two features that result in
quasi-constant axial wave number [26]. The EFB has a constant spiral angle .ζ = 53.7◦

to 230 MeV extraction energy [22, 28]. Hall-probe field measurements are displayed
in Fig. 10.11, together with fields from OPERA computation which are in accord
within a few percent.
Constant Wedge Angle
A spiral EFB ensures constant wedge angle and thus R-independent focusing, while
compensating the axially defocusing effect of a strong field index (Eq. 10.13). A
spiral sector field boundary is defined by
( )
R θ
θ − tan ζ ln
. = constant, i.e., R = R0 exp (10.14)
R0 tan ζ

Note that an . RR0 -homothety, . 2π

N
-rotation orbit similarity results (which reduces to a
simple homothety in a radial lattice as .ζ = 0). The median plane field in a spiral
sector can be written under the form
( )k ( )
R R
. B(R, θ ) = B0 F tan ζ ln −θ (10.15)
R0 R0
( )
wherein .F tan ζ ln RR0 − θ is a . 2π
N
-periodic function of .θ . A convenient model for
the azimuthal modulation assumes a sinusoidal dependence of the field,
( )
R
.F (R, θ ) = 1 + f sin N (tan ζ ln − θ) (10.16)
R0
10.2 Basic Concepts and Formulæ 397

Table 10.2 Design parameters of the RACCAM proton therapy spiral sector scaling FFAG ring.
Some of the parameter values vary with variable operation energy: values given here concern the
extraction energy range .70 to .180 MeV
Injection Extraction
Energy, variable MeV .5.55 → 15 .70 → 180
. Bρ T.m .0.341 → 0.562 .1.231 → 2.030
. Bρextr. /Bρinj. .3.612

.βγ .0.109 → 0.180 .0.393 → 0.648

Lattice type Spiral, scaling
Number of cells .10
.N

Packing factor . pf .0.34

Drift length m .1.15 .1.42

Orbit radius . R m .2.794 .3.460

Orbit excursion m .0.667

(Eq. 10.11)
Wave numbers .2.76;
.ν R ; .ν y .1.55 → 1.60
Transition .2.45
gamma .γtr
Magnet
Type Spiral sector
Sector angle . A deg. .12.24

Spiral angle .ζ deg. .53.7

Scaling index .k .5

Fig. 10.11 Left: measured vertical magnetic field. B Z (θ), along.R = constant arcs across RACCAM
spiral sector dipole [28]. The field on the plateau satisfies the . R k scaling law (Eq. 10.5), lower
(greater) field at lower (greater) energy, greater (lower) dipole gap. Right: field along closed orbits
across the dipole, from OPERA field map computations
398 10 FFAG, Scaling

Transverse Acceptance
Spiral sector scaling FFAG optics features large dynamical transverse acceptance.
As an illustration of that property, the radial dynamical acceptance of RACCAM
spiral sector FFAG ring (Fig. 10.10) is displayed in Fig. 10.12. The latter has been
obtained from raytracing in a theoretical field model built from the EFB geometry
and the . R k dependence of the field, whereas the azimuthal dependence is modeled
using Eq. 10.7 and Enge’s style fall-off (Eq. 14.11) [27].

10.2.3 Longitudinal Motion. Acceleration

Given the orbit length (Eq. 10.8), the revolution period can be written
( ) k+1
1 ( ) k+1
−k
C p β0 p E
T
. rev = = Trev,0 = Trev,0 (10.17)
βc p0 β p0 E0

The momentum compaction and transition .γ write, respectively,

/\C/C 1 / √
α=
. = and γtr = 1/α = 1+k (10.18)
/\p/ p 1+k

Synchrotron Acceleration
Longitudinal focusing and synchronous acceleration in a scaling FFAG proceed as
in synchro-cyclotrons and synchrotrons, as addressed in Sect. 9.2.4.
A practical injection to extraction cycle includes single-bunch or multiturn injec-
tion, RF capture, synchronous acceleration, and single-turn kicker-septum extraction.

Fig. 10.12 Large stability limit (1,000-turn), at various energies, in a spiral sector FFAG. Left:
radial motion; the outer invariants are for pure radial motion, they are several .103 π mm mrad, inner
invariants are the stability limit in the presence of small amplitude axial motion, dynamical accep-
tance decreases to .≈ 103 π mm mrad, an effect of non-linear coupling. Right: 15 MeV case (inner,
elliptical shaped distribution) and 180 MeV case (outer distribution); the dynamical acceptance is
about .600 π mm mrad and .2000 π mm mrad, respectively
10.2 Basic Concepts and Formulæ 399

Fixed-field allows fast cycling, with repetition rate up to hundreds of Hz provided

an appropriate amount of accelerating voltage.
Betatron Damping
In the presence of acceleration (or deceleration) the equations of transverse motion
write [31]
⎧ (γβ), , 1−n
⎪
⎪ x ,,
+ x + x =0
⎨ (γβ) ρ2
. (10.19)
⎪ ,
⎩ y ,, + (γβ) y , +
⎪ n
y=0
(γβ) ρ2

In the adiabatic approximation (damping is slow, compared to betatron motion oscil-

lation) the solutions can be written
[ ({ ) ( { )]
x(s) 1 1
. =/ √ A x exp h x ds + B x exp − h x ds (10.20)
y(s) |h x | βγ y y y y
y

with
⎫
1−n ⎪
⎪
h 2x (s) =− 2 ⎬ [ ] [ ]2
ρ 1 d (γβ), 1 (γβ),
. + + (10.21)
n ⎪
⎪ 2 ds (γβ) 4 (γβ)
h 2y (s) = − 2 ⎭
ρ

and . A x and . B x constants depending upon the initial conditions. Considering that
y y
ρ ∝ R (Eq. 10.4), assuming stable periodic motion, and dropping the .(βγ ), terms
.

in Eq. 10.21 (i.e., h(s) varying slowly), it results from Eq. 10.20 that the transverse
particle oscillations satisfy
√
R 1
. x, y ∝ √ , x ,, y, ∝ √ √ (10.22)
βγ R βγ

thus the damping of betatron oscillations is R-dependent. An invariant ensemble

average results,
[ ]1/2
βγ εr ms = βγ < x 2 >< x ,2 > − < x x , >2
. = constant (10.23)

i.e. transverse emittances of the accelerated motion are damped (or grow if deceler-
ation) according to .εr ms ∝ 1/βγ .
Fast Acceleration
Beyond synchrotron acceleration, alternate methods have been proposed for fast
acceleration in scaling FFAG rings. They are aimed at short-lived particles, high
400 10 FFAG, Scaling

Fig. 10.13 A simulation of serpentine acceleration of a proton bunch from 1.4 to 2.2 GeV (and
deceleration) in a scaling FFAG [32]

Fig. 10.14 Fast stationary

bucket acceleration of a
muon bunch, from 10 to
20 GeV

average current, longitudinal phase rotation. Several of these techniques have been
subject to a proof-of-principle, including quasi-synchronous serpentine accelera-
tion [15] (Fig. 10.13); bucket acceleration [34] (Fig. 10.14); multiple-bunch accelera-
tion [17]; fast longitudinal phase rotation [18]; induction acceleration using a betatron
core [20]; hybrid betatron-synchrotron acceleration [29, 30]; harmonic-jump [16].
10.3 Exercises 401

10.3 Exercises

The following exercises address the two types of scaling FFAG lattices discussed
above: radial sector and spiral sector. Because scaling optics dipoles have a wide
gap, fringe field extent and overlapping may be a concern: the technique described
in Sect. 14.3.3 is used to handle this aspect of the optics.
Note: some of the input data files for these simulations are available in zgoubi
sourceforge repository at
[pathTo]/branches/exemples/book/zgoubiMaterial/FFAG_scaling/.

10.3.1 A 150 MeV, Proton, Radial Sector FFAG

The 150 MeV radial sector FFAG operated at KEK in the early 2000 (Fig. 10.6) is
the subject of the simulations in this series of exercises. Its parameters are given in
Table 10.1, the cell geometry is sketched in Fig. 10.15, the ring geometry and a few
orbits (an outcome of the present exercises) are displayed in Fig. 10.16.

10.1 Field in a Radial Sector Dipole Triplet

Solution 10.1.
The FFAG keyword is based on Eqs. 10.7 and 14.15 to provide an analytical model
of the combined field from N neighboring dipoles, at particle location during the
stepwise integration of motion. The field flutter factor .Fi (R, θ ) (Eq. 14.15) is based
on the fringe field model described in Sect. 14.3.3 (Eq. 14.11).
(a) Using FFAG keyword, and accounting for the magnet and cell parameters of
Table 10.1, produce an input data file for the simulation of a cell.
(b) Produce a graph of the median plane field . B Z (R, θ ) in the dipole triplet. The
keyword OBJET[KOBJ.=1] can be used to generate a dense set of parallel tra-
jectories in the median plane, and OPTIONS[CONSTY.=ON] to force them on
constant radii as they are pushed through the dipoles. Use option FFAG[IL.=2]

Fig. 10.15 Geometry of a

.30
◦ DFD cell. The center of

the ring is at O. F and D are

the focusing and defocusing
sectors of the dipole triplet,
respectively .10.24◦ and
◦
.3.43 . The shadowed .4.75
◦

“E” regions represent a half

of the interval between two
dipole triplets, a region of
.≈200 G stray field [24, 35]
402 10 FFAG, Scaling

Fig. 10.16 A simulation of KEK 150 MeV FFAG ring, and a few closed orbits obtained using the
keyword FFAG. A graph obtained using gnuplot: geometrical data taken from zgoubi.dat, orbit
coordinates read from zgoubi.plt [31]

for a record of step by step trajectory data in zgoubi.plt. Field data can be read
from the latter, to produce a 2D graph of the field . B Z (R, θ ).
(c) While we are here… Using the process in (b) it is possible to generate a mid-plane
field map, on an even 2D meshing, which TOSCA could possibly use and track
through. This requires (i) proper particle sampling for constant ./\R between
trajectory arcs so ray-traced, (ii) proper integration step size FFAG[XPAS] to
cover in an evenly fashion the .30◦ angular sector.
Field data can subsequently be read from zgoubi.plt and re-written in a field map
ascii file with proper formatting for TOSCA to handle.
Work this out, and re-do question (b) to check the identity of the raytracing
outcomes.

10.2 Orbits, Scalloping

Solution 10.2.
The input data file of Exercise 10.1 can be used as a starting point in this exercise.
(a) Compute a scan of the periodic orbits . R(θ ) across the cell, for a few proton
energies ranging in.12 ≤ E ≤ 200 MeV. REBELOTE[IOP.=1, N.=1] can be used
to loop on the energy (by changing the relative particle rigidity. D under OBJET),
preceded by FIT to find the periodic orbit at the energy of concern.
10.3 Exercises 403

Give a graph of these orbits . R(θ ), and on a separate graph the field . B(θ ) along
the orbits. These data can be read from zgoubi.plt, filled using FFAG[IL.=2].
(b) Give a graph of the previous orbits around the ring. Show graphically that these
orbits are similar, check the similarity ratio.
(c) By tracking, show that orbit excursion over an energy range .12 ≤ E ≤ 200 MeV
(average radius spans from . Rin j to . Rxtr ), satisfies Eqs. 10.8, 10.11. Particle coor-
dinates at some azimuth along the ring can be logged in that aim in zgoubi.res
using FAISCEAU (a linux “grep” can then grab them for plotting), or in an
ancillary zgoubi.fai file using FAISTORE.
(d) Evaluate the orbit scalloping, i.e., the maximum value of .|R(θ ) − R|/R. Give a
graph of the latter as a function of energy.

10.3 Zero-Chromaticity
Solution 10.3.
This exercise investigates the momentum dependence of the wave numbers.

(a) Compute and give a graph of the momentum dependence of the radial and axial
wave numbers in the 12-cell ring (Fig. 10.6). Use for that either one of the
following two methods to obtain the wave number values:
(i) From the cell transport matrix, using MATRIX[IORD.=1, IFOC.=11].
REBELOTE[IOP.=1, N.=1] can be used in that case to repeat on momentum
values.
(ii) from Fourier analysis of small amplitude motion.
Compare the results with theory (Eq. 10.13).
(b) It can be observed that the radial wave number is constant with momentum/orbit
radius. R, this is expected from the scaling law (Eq. 10.5); however the axial wave
number is R-dependent.
In the field model, using FFAG[.κ] EFB parameter, introduce an R-dependence
of the gap height (Eq. 10.6): this is equivalent to introducing an R-dependence of
the fringe field extent, or equivalently of the field form factor .F (R, θ ) (Eq. 10.7),
proper to change the . R-dependence of the axial focusing. Find the value of .κ
which minimizes the change of .ν y over the energy interval .12 ≤ E ≤ 150 MeV,
provide a simulation to show the efficiency of the method.
(c) Compute the value of the momentum compaction and transition.γtr at two sample
energies, 12 and 150 MeV. TWISS can be used for that, with OBJET[KOBJ.=5].
Check their relationship to the radial wave number.

10.4 Beam Envelopes; Phase Space

Solution 10.4.

(a) Produce a graph of the trajectories of a beam bundle across the cell, at 12 and
150 MeV. Take initial coordinates evenly distributed on initial paraxial invariants.
OBJET[KOBJ.=8] can be used to define that set of particles.
404 10 FFAG, Scaling

(b) Perform single particle tracking, over many turns, using REBELOTE. Consider
two cases, separately: paraxial motion, and large excursion motion. Show that
large excursion phase space motion features non-linear coupling.

10.5 Acceleration. Transverse Betatron Damping

Solution 10.5.
(a) Produce a simulation of the transverse and longitudinal motions of a par-
ticle taken on a small initial invariant, over a .10 → 150 MeV acceleration
cycle in the 12-cell ring. Assume the following RF parameters: peak voltage
◦
. V̂ = 40 kVolts, synchronous phase .φs = 20 , harmonic h .= 1. Acceleration uses

CAVITE[IOPT.=6], which imposes defining the particle type, with PARTICUL;

multiturn is obtained using REBELOTE. SCALING[IOPT.=−2] is used to read
RF program data for CAVITE[IOPT.=6], from an ancillary file (with default file
name zgoubi.freqLaw.In).
(b) Show graphically that the transverse betatron oscillation damping satisfies the
. R-dependence of Eq. 10.22.
(c) Accelerate a bunch of a few tens of particles. Check the beam emittance damping
of Eq. 10.23.

10.3.2 RACCAM Proton Therapy Spiral Sector FFAG

This series of exercises is based on the 180 MeV spiral sector FFAG design of
Fig. 10.10. The parameters of concern are given in Table 10.2 [22, 26–28]. The cell
geometry is sketched in Fig. 10.17.

Fig. 10.17 A sketch of

RACCAM spiral sector
dipole and .2π/10 cell. O is
the center of the ring and the
EFBs form a sector angle . A.
The reference orbit is not
circular, the bending radius
is not constant along the
trajectory over the .2π/N
arc—a line of constant field
is an arc of a circle, centered
at . O
10.3 Exercises 405

10.6 Field in a Spiral Sector Dipole

Solution 10.6.
The FFAG-SPI keyword is based on Eq. 10.15 to generate the field from a spiral sector
dipole (or several sectors side-by-side within an AT angular extent [33, Sect. FFAG-
SPI]) at particle location, while motion proceeds across the magnet. FFAG-SPI has
provision for the modeling of the azimuthal form factor .Fi (R, θ ), and the overlap-
ping of neighboring fringe fall-offs, based on the method described in Sect. 14.3.3
(Fig. 10.18).
(a) Using FFAG-SPI keyword, produce a graph of the median plane field . B Z (R, θ )
in RACCAM spiral sector dipole. OBJET[KOBJ.=1] can be used to generate a
trajectory sample, and OPTIONS[CONSTY.=ON] to force these trajectories on
constant radii. Option FFAG-SPI[IL.=2] will log in zgoubi.plt the step by step
trajectory field data; the latter can subsequently be plotted.
(b) While we are here… Using the process in (a) it is possible to generate a mid-
plane field map, on an even 2D meshing, which TOSCA[. I Z > 1, M O D = 25]
could possibly use and track through. This requires (i) proper particle sampling
in OBJET for constant ./\R between trajectory arcs so ray-traced, (ii) proper
integration step size FFAG[XPAS] to cover in an evenly fashion the .36◦ angular
sector.
Field data can subsequently be read from zgoubi.plt and logged in a field map
file with proper formatting for TOSCA[. I Z > 1, M O D = 25] to handle [33,
cf. TOSCA].
Work this out, and re-do question (b) using TOSCA, to check the accord of the
raytracing outcomes with the FFAG-SPI case.

Fig. 10.18 A simulation of

RACCAM FFAG ring,
including a few orbits, using
the keyword FFAG-SPI. A
graph obtained using
gnuplot, geometrical data
taken from zgoubi.dat, and
particle data read from
zgoubi.plt
406 10 FFAG, Scaling

(c) CYCLOTRON keyword could be used as well: it allows some sophistication

in field modeling compared to using FFAG-SPI, such as accounting for an R-
dependence of the geometrical field index .k (a capability which is used in Exer-
cise 4.6—Relativistic Cyclotron Chapter, to adjust the isochronism), and of the
fringe field extent, via an R-dependent gap shape index .κ(R) (Eq. 10.6).
Move the FFAG modeling of (a) to CYCLOTRON keyword. Try some radial
dependence of both .k and fringe extent, with the constraint of maintaining constant
radial and axial tunes over the spiral orbit radial excursion. FIT can be used for this
optimization of tune constancy.

10.7 Orbits, Scalloping

Solution 10.7.
Characterizing the focusing properties of the lattice (say, over the radial span
of the accelerated orbit) first requires finding the periodic orbits. The radius—or
momentum—dependence of optical functions may then be found (Exercise 10.9), as
well as the radius dependence of time of flight for further acceleration
(Exercise 10.13), etc.
(a) Compute a scan of the periodic orbits . R(θ ) across the cell, for a few proton
energies ranging in .15 ≤ E ≤ 180 MeV. REBELOTE[IOPT.=1] can be used to
loop on the energy (by changing the relative particle rigidity . D under OBJET),
preceded by FIT to find the periodic orbit at the energy of concern.
Give a graph of these orbits. R(θ ), and on a separate graph the field. B(θ ) along the
orbits. These data can be logged in zgoubi.plt during ray-tracing, using FFAG-
SPI[IL.=2].
(b) Show graphically the homothety-rotation of the orbits.
(c) By tracking, show that orbit excursion over an energy range .15 ≤ E ≤ 180 MeV
(average radius spans from . Rin j to . Rxtr ), satisfies Eqs. 10.8 and 10.11.

10.8 Zero-Chromaticity
Solution 10.8.

(a) Compute and give a graph of the momentum dependence of the radial and axial
wave numbers in the 10-cell ring (Fig. 10.10). Use for that either one of the
following two methods to obtain the wave number values:
(i) from the cell transport matrix,
(ii) from Fourier analysis of paraxial motion.
Compare with expectations (Eq. 10.13).
(b) It can be observed that the radial wave number is constant with momentum,
or equivalently with the orbit radius . R, this is expected from the scaling law
(Eq. 10.5). However the axial wave number is R-dependent. Explain why.
(c) In the field model, introduce an R-dependence of the gap height following
Eq. 10.6: this is equivalent to introducing an R-dependence of the fringe field
extent, or equivalently of the field form factor .F (R, θ ) (Eq. 10.7), proper to
10.3 Exercises 407

change the . R-dependence of the axial focusing. Using the FIT procedure, com-
pute the value of .κ which minimizes the change of .ν y over the energy interval
.15 < E < 180 MeV.
(d) Compute the value of the momentum compaction and transition .γtr , at 15 and
180 MeV. Check their relationship to the radial wave number.

10.9 Beam Envelopes, Optical Functions

Solution 10.9.
From single-particle multiturn tracking, produce graphs of radial and axial excursion
across the cell, at 15 and 180 MeV. From these multiturn data, derive the betatron
function amplitudes.

10.10 Periodic Stability Domain

Solution 10.10.
Vary the scaling index .k and spiral angle .ζ of the spiral dipole, in FFAG-SPI:

– produce a two-dimensional .(ν R , ν y ) wave number scan diagram, covering the

motion stability area resulting from varying .k and .ζ .
– produce the corresponding .(k, ζ ) stability limit diagram.

10.11 Motion Stability Limit

Solution 10.11.
Tracking single particle radial motion in the OPERA field map yields at stability
limit the phase space portrait of Fig. 10.12 [26]. Re-produce a similar phase space
graph at stability limit, using the analytical field model FFAG-SPI.

10.12 Dynamic Aperture

Solution 10.12.
Extend the previous stability limit search (Exercise 10.11) to producing the dynamic
aperture in (Y, Z) space, at 15, 57 and 180 MeV.

10.13 Acceleration. Transverse Betatron Damping

Solution 10.13.
Produce a simulation of a .15 → 180 MeV acceleration cycle in RACCAM ring, for
a single particle with paraxial radial and axial motions. Take an acceleration rate of
10 kVolts per turn. Acceleration uses CAVITE[IOPT.=6], which imposes defining
the particle type, with PARTICUL; multiturn is obtained using REBELOTE[K.=99].
SCALING[NFAM=CAVITE[NT.=−2]] takes care of reading the RF program from
zgoubi.freqLaw.In.
Show the betatron damping, graphically, check Eqs. 10.22, 10.23.
408 10 FFAG, Scaling

10.3.3 FFAG Acceleration Methods

Regarding the lattice, the following three exercises are based on a similar radial
sector triplet FFAG to that studied in detail in the Sect. 10.3.1 exercise series. Thus
earlier simulation input data files can be resorted here, and will only require minor
adaptations.
Regarding beam acceleration, the input data files and methods developed in Exer-
cises 10.13, 10.5 can be used to set up the present acceleration simulation input
data.

10.14 Hybrid Acceleration

Solution 10.14.
Produce a simulation of hybrid acceleration in the .35 keV.→ 7 MeV, .C4+ , FFAG
injector addressed in Ref. [29, Slides 17–18]. It is suggested to proceed with staged
simulations in the following order:
(a) Build the input data file for a .k = 0.7, radial sector DFD cell, and subsequently
for an 8-cell ring. Check the paraxial parameters, using OBJET[KOBJ.=5] and
MATRIX[IORD.=1, IFOC.=11] or TWISS, or OBJET[KOBJ.=6] and MATRIX
[IORD.=2, IFOC.=11]. The methods of the exercises in Sect. 10.3.1 can be used
to construct the cell.
(b) Add acceleration, using a single, 5 kV RF cavity, simulated using CAVITE;
(c) add betatron-style acceleration, using CAVITE[IOPT]—proper choice of IOPT
option is to be determined.

10.15 Bucket Acceleration

Solution 10.15.
Produce a simulation of bucket acceleration of a short-lived muon bunch (cf.
Fig. 10.14), from 3.6 to 12.6 GeV, following Ref. [16, pp. 4507–4508]. It is sug-
gested to proceed with staged simulations in the following order:
(a) Set up a 225-cell,.k = 1390, DFD ring. Check its parameters. The methods of the
radial lattice exercises (Sect. 10.3.1) can be used to construct the cell. Check the
paraxial parameters, using OBJET[KOBJ.=5] and MATRIX[IORD.=1,
IFOC.=11] or TWISS, or OBJET[KOBJ.=6] and MATRIX[IORD.=2,
IFOC.=11].
(b) Add acceleration, 1.8 GV per turn, using 225 (one per cell), 8 MV, 200 MHz RF
cavities, harmonic .h = 675. Re-produce Figs. 5, 6 of Ref. [16, p. 4508].
(c) Add Monte Carlo in-flight decay (MCDESINT keyword): find the muon sur-
vival rate over the acceleration cycle. Check against theoretical expectations (cf.
Sect. 12.2.3).
10.4 Solutions of Exercises of This Chapter: FFAG, Scaling 409

Table 10.3 Parameters of K. Yamakawa’s 1.1 GeV quasi-isochronous scaling FFAG ring [32]
Lattice FDF triplet
Number of cells To be determined.a
k-value 1.45
Transition energy [MeV] 530
Equivalent mean radius at 200 MeV [m] 3
Equivalent mean radius at 1 GeV [m] 5.9
Stationary kinetic energy below transition [MeV] 360
rf voltage [MV/turn] 15 (.h = 1)
rf frequency [MHz] 9.6 (.h = 1)
a
. As part of the exercise

10.16 Serpentine Acceleration

Solution 10.16.
Produce a simulation of 0.38–1.1 GeV proton beam serpentine acceleration, in 10
turns, following the hypotheses found in Ref. [32] for an MW power Accelerator
Driven System. Parameters are recalled in Table 10.3.
It is suggested to proceed with staged simulations in the following sequence:
(a) Set up a data file for a ring, according to parameters in Table 10.3. The meth-
ods of the radial lattice exercises (Sect. 10.3.1) can be used to construct the
cell. Check consistency with Table 10.3 data. Produce the paraxial parame-
ters using OBJET[KOBJ.=5] and MATRIX[IORD.=1, IFOC.=11] or TWISS,
or OBJET[KOBJ.=6] and MATRIX[IORD.=2, IFOC.=11].
(b) Add acceleration, using a single, 60 MV RF cavity, harmonic .h = 10.
CAVITE[IOPT.=7] can be used. Produce the longitudinal phase space of an
acceleration cycle (similar to Fig. 10.13).

10.4 Solutions of Exercises of This Chapter: FFAG, Scaling

10.4.1 A 150 MeV, Proton, Radial Sector FFAG

10.1 Field in a Radial Sector Dipole Triplet

(a) An input data file to simulate a 150 MeV scaling FFAG cell.
The input data file, including comments for guidance, is given in Table 10.4. Geome-
try and field data under FFAG are from Table 10.1, possibly slightly varied according
to the purposes of subsequent questions.
(b) A graph of . B Z (R, θ )|Z=0 in a radial sector scaling FFAG dipole.
In view of generating the mid-plane field from raytracing outcomes, a set of 25 tra-
jectories is launched, with initial coordinates .Y0 , T0 , Z 0 , P0 and relative momentum
. D0 = p/ pref , defined using OBJET[KOBJ=1] (Table 10.4). Initial radii .Y0 are evenly
spaced over the useful field region, namely
410 10 FFAG, Scaling

Table 10.4 Simulation input data file SFFAGCell.inc: 150 MeV KEK FFAG dipole triplet cell,
a .30◦ sector. The FFAG keyword allows defining up to 5 independent dipoles in that . AT =
30◦ angular sector, only three are needed for the present DFD triplet. OPTIONS[CONSTY
ON] allows to raytrace a set of trajectories on constant radii. This file also defines the seg-
ment #S_SFFAG150Cell to #E_SFFAG150Cell, for use in INCLUDEs in subsequent exer-
cises. The step size value . X P AS = 3.0079078598 cm here is for field map fabrication,
change to ∼1 cm for multiturn tracking Note: this file is available in zgoubi source-
forge repository at https://sourceforge.net/p/zgoubi/code/HEAD/tree/branches/exemples/book/
zgoubiMaterial/FFAG_scaling/radialSectorTriplet_KEK/accelerationCycle
10.4 Solutions of Exercises of This Chapter: FFAG, Scaling 411

Y : r1 → r25 , step /\r

. 0 with r1 = 446 cm, r25 = 518cm, /\r = 3 cm

.This.Y0 range results from Eq. 10.8, given.k = 7.6 (Table 10.1) and. R0 = 540 cm (the
choice of . R0 is arbitrary). As a mid-plane field is desired, axial motion is taken null,
. Z 0 = 0 and . P0 = 0. By means of OPTIONS[CONSTY ON] each particle is forced
to maintain constant radius throughout the dipole. Note that a 3D map of the vector
_
field instead, . B(R, θ, Z ), can be generated if desired, by adding a . Z 0 sampling, as
CONSTY also forces Z to maintain its initial value . Z 0 .
The integration step size along the reference arc . R0 is ./\s = 3 cm, resulting in 94
steps over

θ : θ1 → θ95 , step /\θ = /\s/R0 , with θ1 = 0, θ95 = 30◦ , /\θ = 0.31915◦

.

.with the.30◦ triplet sector opening including half-drifts on both sides to make up a cell
(Fig. 10.15). In that manner, during the stepwise raytracing process, the mid-plane
field . B Z (R, θ ) is computed at particle locations which are made to coincide with the
. Nr × Nθ = 25 × 95 nodes of a 2D meshing.
Note: although not necessary as far as the present question (b) is concerned, this
meshing has been tailored to be uniform, and to exactly cover the .30◦ sector, in view
of the next question.
The magnetic field vector experienced along the trajectory across the dipoles is
part of the particle data logged in zgoubi.plt during the stepwise integration, as an
effect of the flag FFAG[IL.=2] (Table 10.4). A graph of . B Z (R, θ ) data so obtained,
read from zgoubi.plt, is given in Fig. 10.19.

Fig. 10.19 Using the analytical field model provided by the FFAG keyword (Eq. 10.7), the 2D
mid-plane field of KEK 150 MeV DFD dipole triplet is produced (left), over a uniform 2D polar
meshing (right) defined by OBJET[.δY ] particle sampling and by the step size FFAG[XPAS] (./\θ
sampling)
412 10 FFAG, Scaling

(c) Generating a 2D mid-plane field map (which TOSCA could possibly handle).
A straightforward way to generate a field map for possible use by TOSCA is to get
field data on a uniform 2D mesh from the raytracing, using appropriate integration
step size and particle sampling.
This has been accounted for in the input data file for the previous question,
Table 10.4: the radial increment ./\R was defined to be constant using OBJET
[KOBJ.=1]. The angular increment ./\θ = (π/6) /94[steps] is constant by defini-
tion; what matters, and accounted for in (b), is ensuring that the last step (down-
stream boundary of the mesh) is on the exit border of the .30◦ sector: this is
ensured taking an integration step size . X P AS = R0 [540 cm] × (π/6) /94[steps] =
3.0079078598 cm.
From this, it results evenly distributed .(Ri,j , θi,j ) particle locations during the
raytracing; step-by-step particle data logged in zgoubi.plt include coordinates and
the field values . B Z (Ri,j , θi,j , Z = 0), they are read to be re-written in an ASCII file
with proper formatting for TOSCA to handle and track through. The appropriate
formatting can be found in [33, Table 1].
Questions 3.1, 4.1 and their solutions may be resorted to in working out the details
of the present question. This is left to the reader.
10.2 Orbits, Scalloping
(a) Periodic orbits.
Based on REBELOTE do-loop, the input data file in Table 10.5 produces 10 closed
orbits consecutively (FIT finds them, one after the other) for as many differ-
ent momenta (REBELOTE repeats the sequence for 10 different momenta) rang-
ing in (relative to . pref = 551.345 MeV/c, 150 MeV proton) . p/ pref : 0.2730426 →
1.168858 (12–200 MeV).
The input data file ends with a SYSTEM command which results in the two graphs
displayed in Fig. 10.20.
(b) Homothetic orbits.
Homothetic orbits around the ring, obtained from the previous question, read from
zgoubi.plt, are plotted in Fig. 10.21.
It is found from Fig. 10.20 that the scalloping is about .0.2/5.4 ≈ 3.7% for the
high energy closed orbit, about .0.2/4.5 ≈ 4.4% for the low energy closed orbit.
The similarity ratio . ρR (Bρ) is expected to be close to constant, within a few %, so
justifying the oft-met assumption that, at all azimuth, . R ≈ C/2π . It can be computed
for these 10 different rigidities: orbit radius . R and field . B Z in the region .θ ≈ 15◦ are
read from the step by step . R(θ ), . B Z (θ ) data logged in zgoubi.plt, the latter yielding
◦ ◦
.ρ(θ = 15 ) = Bρ/B(θ = 15 ). Results are displayed in Fig. 10.22.

(c) Orbit excursion.

Figure 10.23 displays the numerical and theoretical (Eq. 10.8) values of the average
orbit radius, they appear in good accord. It results from this that Eq. 10.11 is satisfied,
with similar accuracy, .≈ 1%.
10.4 Solutions of Exercises of This Chapter: FFAG, Scaling 413

Table 10.5 Simulation data file to find the closed orbit for a series of different momenta. The
INCLUDE grabs the FFAG dipole triplet segment defined in Table 10.4

(d) Orbit scalloping

A gnuplot script can plot the scalloping .|R(θ ) − R|/R. . R(θ ) is read from zgoubi.plt
(column 10 therein) as in (b). The value of the average orbit radius . R = C/2π can be
formulated using Eq. 10.11 with . p = q Bρ = q × D × B O R O, with D and BORO
both read from zgoubi.plt (columns 2 and 40, respectively [33, Sect. 8.3]).
414 10 FFAG, Scaling

Fig. 10.20 Left: orbit scalloping across the AT=30.◦ angular extent encompassed in FFAG keyword
simulation, for 10 different proton energies ranging in 12–200 MeV (from bottom, smaller radius,
to top, greater radius). Right: field experienced along these orbits, increasing with radius

Fig. 10.21 Ten closed orbits, from 12 to 200 MeV, around the 12-cell radial sector FFAG ring. A
graph obtained using gnuplot, all necessary data read from zgoubi.plt
10.4 Solutions of Exercises of This Chapter: FFAG, Scaling 415

Fig. 10.22 Left: curvature radius .ρ(θ) across the focusing dipole of the DFD triplet, for ten closed
orbits in the energy range .12 < E < 200 MeV. Right: . R/ρ(θ) in the central region of the focusing
dipole; the variation is .≈ ±0.1/4.75 ≈ ±2%

Fig. 10.23 Dependence of the average closed orbit radius . R = C/2π on the relative momentum,
from 12 to 200 MeV. Markers are from one-turn raytracing. The solid line is from theory, for com-
parison, after Eq. 10.8 taken for.k = 7.6 (Table 10.1) and reference momentum. pref = 551.3 MeV/c
(150 MeV, radius . R0 = 5.4 m). They only differ by .≈ 1%

10.3 Zero-Chromaticity

(a) Momentum dependence of tunes.

MATRIX is used to compute the wave numbers, REBELOTE[K.=0;IOPT.=1] is used

to repeat with a different momentum, the input data file is given in Table 10.6.
OBJET[KOBJ.=5] defines a set of 13 particles with proper initial coordinate
sampling for matrix computation, by MATRIX. Prior to matrix computation, the
momentum-dependent closed orbit is found by FIT. REBELOTE then changes
the relative momentum D in OBJET, and repeats the sequence. The command
416 10 FFAG, Scaling

Table 10.6 Simulation input data file: compute the first order transport matrix of the DFD cell, for
a series of momenta. Prior to matrix computation, the closed orbit is found by FIT. Next to that,
REBELOTE repeats for a new momentum value. The INCLUDE grabs the dipole triplet segment
of Table 10.4

MATRIX[PRINT] logs the transport coefficients into zgoubi.MATRIX.out, together

with the beam matrix and tunes which are obtained from the hypothesis of periodicity
(MATRIX[IFOC.=11]), namely by identifying (Sect. 14.5.2)

.[Ti j ] = I cos μ + J sin μ

Outcomes are plotted in Fig. 10.24. The radial tune .ν R is constant as expected from
.

the zero-chromaticity resulting from the scaling law (momentum-independent index,

Eq. 10.3). Such is not the case for the axial tune, .ν Z , this is due to the first order effect
of fringe field extent on the axial focusing (see Sect. 14.4.1): fringe extent varies with
orbit radius in correlation with gap height in the “gap shaping” hypothesis (Eq. 10.6;
smaller gap at greater radius/greater energy).
10.4 Solutions of Exercises of This Chapter: FFAG, Scaling 417

Fig. 10.24 Radial (.ν R ) and axial (.ν Z ) tunes (left vertical scale) of the 12-cell ring, as a function
of relative momentum (. p/ pref = 1 for 150 MeV). The penalty values (scattered squares, right
vertical scale) monitor the FIT runs: a small penalty indicates convergence of FIT. Left: case of
( )3
decreasing gap height with radius (equivalent to a decreasing fringe extent): .g(R) = g0 RR0 .
Right: case of linear( increase
) of gap height with radius (equivalent to a linear increase of fringe
extent): .g(R) = g0 RR0 . .ν Z is now about constant over the momentum span

(b) Momentum-dependent axial tune.

The variation of the axial tune with momentum, as observed in (a), is due to the
fringe field extent decreasing with radius, following in that the gap height which
induces the scaling field law . B(R) ∝ R k (Eq. 10.5): as a matter of fact the gap shape
index value in the present radial sector model is .κ = 3 (Table 10.4), resulting in a
gap height (Eq. 10.4)
( )3
R0
. g(R) = g0
R

The axial tune can be made constant using instead a gap law (Eq. 10.6)
.

( )
R
. g(R) = g0 (κ = −1)
R0

i.e., gap height proportional to momentum/R [22, 26]. The simulation is obtained by
.
changing the lines of concern under FFAG keyword (Table 10.4), namely

change “6.3 3. ! EFB [etc.]”

. to “6.3 − 1. ! EFB [etc.]”

at the 6 EFBs. The resulting tunes are displayed in Fig. 10.24: .ν R is marginally
affected, whereas .ν Z is now about constant.
A FIT procedure can be further attempted, in order to try and improve things,
proceeding in the following way:

– vary .κ
– constrain .ν Z = constant at a few energies in the 12–150 MeV range.
418 10 FFAG, Scaling

Table 10.7 Simulation input data file: TWISS command, to obtain beam matrix, momentum com-
paction, chromaticities, etc. The initial reference coordinates, under OBJET, are for 12 MeV; refer-
ence coordinates for 150 MeV (to substitute to the previous ones) have been added as a comment.
TWISS is an implicit do-loop, it proceeds in 3 stages: it first computes tunes of an on-momentum
particle, then for .±δp/ p off-momentum particles; at each stage, FIT ensures that the reference
particle (1st particle of the 11-set) is on the closed orbit

However, as expected from theory, it is found that the constraint is already fairly
satisfied with .κ = −1.

(d) Momentum compaction and transition .γtr

TWISS keyword is used concurrently with OBJET[KOBJ.=5], to compute various

first and second order optical parameters including chromaticities.
MATRIX[IORD.=2, IFOC.=11] concurrently with OBJET[KOBJ.=6] can be used
as well. A typical input file is given in Table 10.7. Results are given in Table 10.8.
It is expected for the momentum compaction to satisfy .α = /\C C
/ /\p
p
= 1/(1 + k)
(Eq. 10.18). In the present design.k = 7.6 in all three dipoles of the triplet (Table 10.4),
yielding .α = 0.11628. This approximation appears valid at high energy where ray-
tracing yields .α = 0.11620 (Table 10.7), however it is not the case at low energy
where the numerical integration yields .α = 0.42528. This discrepancy might result
from the greater flutter at smaller orbit radius (shorter magnet body compared to
fringe field extent)—further investigation is left to the reader.
10.4 Solutions of Exercises of This Chapter: FFAG, Scaling 419

Table 10.8 Outcomes of TWISS computation out of zgoubi.res execution listing, including beam
matrix, tunes, momentum compaction, chromaticities

10.4 Beam Envelopes; Phase Space

(a) Motion envelopes.

There is various possibilities to get the beam envelopes along the cell. One consists
in raytracing a few particles with initial coordinates taken on an ellipse. Another
method tracks a single particle, for a few tens of turns. A third possibility consists in
pushing the initial beam matrix.σ (s0 ) through the cell, using.σ (s) = T (s ← s0 )σ (s0 )
T̃ (s ← s0 ) (Sect. 14.5.2), by computing .T (s ← s0 ) from the stepwise particle coor-
dinates in the option OBJET[KOBJ.=5]—a tool to push.σ (s) can be found in zgoubi
toolbox [36]. √
In any case, linear envelopes, with maximal excursion . ( επY βY (s)) and
√ εZ
. ( β (s)), require paraxial motion.
π Z
The first method is retained, here. Set IL.=2 under FFAG to have particle data
logged, step by step, in zgoubi.plt. Graphs of the trajectories of the beam bundle
across the cell, at 12 and 150 MeV are given in Fig. 10.25.
420 10 FFAG, Scaling

Fig. 10.25 Radial (left) and axial (right) beam bundle trajectories across the FFAG triplet cell; top
row: 12 MeV, bottom row: 150 MeV. Graphs obtained using zpop: menu 7; 1/1 to open zgoubi.plt;
2/[8, 2] to select .Y (radius) versus . X (azimuthal angle) [or 2/[8, 4] to select . Z (axial coordinate)];
7 to plot

Table 10.9 Simulation input data file: proper OBJET, using KOBJ.=8, to define a beam bundle by
initial reference orbit coordinates, and Courant invariant values

Particle trajectories result from initial coordinates taken on a small invariant

value (an ellipse), in both radial and axial planes, namely, .εY /π = ε Z /π = 0.1 mm
(Table 10.9). The definition of the initial coordinates on an ellipse uses the keyword
OBJET[KOBJ.=8]. The ellipse parameters, .αY,Z , .βY,Z , are taken from TWISS out-
comes of the previous exercise (Table 10.8).
Note that the rather large initial invariant values (.0.1π mm) result in the two
motions to be slightly coupled (in the presence of non-zero axial motion), the initial
elliptical invariant is not strictly preserved during the propagation, some smear shows,
see next question.
10.4 Solutions of Exercises of This Chapter: FFAG, Scaling 421

Fig. 10.26 Multiturn tracking: radial (left) and axial (right) phase space motion, observed at the
end (middle of the drift) of an FFAG triplet cell; top row: 12 MeV, bottom row: 150 MeV. The central
quasi-elliptical motion is for .εY /π = ε Z /π = 1 µm, the outer motion, distorted and coupled, is for
.εY /π = ε Z /π = 0.1 mm. Graphs obtained using zpop: menu 7; 1/5 to open zgoubi.fai; 2/[2, 3] to
select .T versus .Y (or, 2/[4, 5] to select . P versus . Z ); 7 to plot

(b) Multiturn tracking, phase space.

The definition of the initial coordinates of the particle to be tracked uses the keyword
OBJET[KOBJ.=8] as in the previous method. However, considering the input data
in Table 10.9,
.change “30 30 1” to “1 1 1”

Multiturn tracking reveals that.0.1 mm motion invariants are large enough that (i) they
.

are distorted by field non-linearities (compared to ellipses in the case of parax-

ial motion), and (ii) Y and Z motions feature non-linear coupling. This shows in
Fig. 10.26 which displays phase space motion in the two cases of initial coordi-
nates taken on an .εY /π = ε Z /π = 0.1 mm ellipse and on an .εY /π = ε Z /π = 1 µm
ellipse. In the first case the motion is not elliptical, whereas in the second case, parax-
ial conditions, the phase space portrait is an ellipse, with no visible coupling (a thin
ellipse, in both planes).

10.5 Acceleration. Transverse Betatron Damping

(a) Acceleration cycle in an FFAG ring, using an RF program.
A typical prior check: it is a good idea to first validate the input data file to be
used for acceleration, by producing and checking a few closed orbits, or one-turn
422 10 FFAG, Scaling

Table 10.10 Simulation input data file: checking the file set up for acceleration. This data file is
derived from the acceleration input file in Table 10.12, and provides periodic matrix computation
at 10 MeV and 200 MeV, combining OBJET[KOBJ.=5] (to allow computation by MATRIX), FIT
(find the orbit) and REBELOTE (repeat for an additional momentum)

beam matrices. An input data file in that aim is given in Table 10.10. Excerpts from
zgoubi.res execution listing so obtained are given in Table 10.11, they show periodic
orbits at expected radii, respectively .Y0 = 440.5 cm (10 MeV) and .Y0 = 526.9 cm
(200 MeV), and expected beam matrices (compare to Table 10.8).
Following these preliminary checks, an input data file set for acceleration from 10
to 200 MeV is derived, Table 10.12. Note the “12 *” under INCLUDE, for 12 cells.
The cell, a dipole triplet with half-drifts on both sides, is as in Table 10.4, yet making
sure for the values of the following three parameters:
(i) IL is set to 0, to inhibit output to zgoubi.plt as this saves on computing time—
another possibility is to use OPTIONS[.plt 0]
(ii) the integration step size is set to ./\s = 1 cm, for accuracy over the 13,000 turn
acceleration cycle
(iii) FFAG allows a few methods for the numerical integration, KIRD.=0 was used in
Table 10.4, whereas the method retained here instead, for a change (an interest-
ing exercise would consist in comparing the outcomes from the two methods),
is KIRD.=2; as a consequence

“0 2 ! Field computation analytic 2nd order (take KIRD = 2, 25 or 4 etc.”

.

in Table 10.4, is changed to (comments, beyond the exclamation mark, have no

effect)

“2 10. ! Field computation method: 3 ∗ 3 flying grid, 2nd deg interpolation.”

.
10.4 Solutions of Exercises of This Chapter: FFAG, Scaling 423

Table 10.11 Output file: checking the file set up for acceleration. This table shows excerpts from
zgoubi.res execution listing, following from the input data file in Table 10.10, namely, the closed
orbit, matrix and tunes for 10 MeV (relative momentum D.=0.2491, top part) and for 200 MeV
(D.=1.1689, bottom part). Particle 1 is the reference particle for the computation of the transport
matrix from which the beam matrix is deduced

The consequence is that the field and derivatives [33, Sects. 1.2 and 1.2.1] are com-
puted using a .3 × 3 node flying grid technique (that is what ‘2’ stands for, ‘10’ stands
for the grid mesh size, taken equal to./\s/10), whereas in the previous exercises, given
KIRD.=0, field and derivatives are computed from (hard-coded) analytical expres-
sions [33, Part B, FFAG] [35].
Setting up the acceleration, now:

– PARTICUL[PROTON] is necessary as CAVITE is used: it allows converting

energy change in rigidity change (zgoubi pushes particles using rigidity),
– SCALING[IOPT.=-2], akin to a “power supply rack”, provides the RF program
to CAVITE, by reading it from the ancillary file zgoubi.freqLaw.In, see below,
– CAVITE[IOPT.=6] boosts the particle(s) at each pass, following that pre-defined
RF program,
– REBELOTE sends the execution pointer back to the top of the input data file, for
multiturn tracking. The number of turns results from a peak voltage .V̂ = 40 kV
and synchronous phase .20◦ (Table 10.12), the 10 to 200 MeV acceleration range
is covered in .(200 − 10) × 103 /[40 × sin(20◦ )] ≈ 13900 turns.

For simplicity the RF program is limited in the present case to the turn depen-
dence of RF frequency (peak voltage and synchronous phase maintained constant).
424 10 FFAG, Scaling

Table 10.12 Simulation input data file: proton acceleration from 10 MeV to about 200 MeV
using zgoubi.freqLaw.In RF program file. Note: the step size must be in the cen-
timeter range for multiturn accuracy, in such non-linear field (make sure to substi-
tute “1.” (cm) to “3.0079078598”, in the INCLUDE file SFFAGCell.inc, Table 10.4).
Note: this file, and all INCLUDE files it resorts to, are available in zgoubi source-
forge repository at https://sourceforge.net/p/zgoubi/code/HEAD/tree/branches/exemples/book/
zgoubiMaterial/FFAG_scaling/radialSectorTriplet_KEK/accelerationCycle

Elaborating zgoubi.freqLaw.In (essentially two columns: RF frequency versus turn

number) requires the following steps:
(i) run a search for closed orbits for a few tens of energies in the range 10–200 MeV.
The corresponding data file is given in Table 10.13. Time of flight, derived from
path length and particle velocity, is part of the outcome of orbit computation
as PARTICUL provides necessary particle data in that aim. That yields the
turn dependence of RF frequency, namely, . f RF = h × f rev = v/C (.h = 1, here).
Details below. Note that in the absence of orbit defects or other tailored bumps,
the frequency law may be obtained, more simply, from the theoretical orbit length
1
.C( p) = C0 ( p/ p0 ) k+1 (Eq. 10.8);
(ii) store these turn-frequency data, properly formatted, in zgoubi.freqLaw.In
(Table 10.14). zgoubi.freqLaw.In does not need to contain every turn, zgoubi
will interpolate from the set of values provided.
(b) Transverse motion.

The acceleration simulation file is that of Table 10.12. Longitudinal and transverse
motion samples are displayed in Fig. 10.27. The integration step size is ./\s = 1 cm
in these simulations. Taking KIRD.=0 instead (see remarks above), all the rest
unchanged, would result in a marginal difference.
10.4 Solutions of Exercises of This Chapter: FFAG, Scaling 425

Table 10.13 Simulation input data file: search closed orbits for a few tens of energies in the range
10–200 MeV, for fabrication of zgoubi.freqLaw.In RF program file

Table 10.14 Top and bottom parts of the RF program file zgoubi.freqLaw.In, as read by zgoubi when
using CAVITE[IOPT.=6]. Zgoubi actually only requires turn number, column 1, and revolution
time which is computed from the cumulated time-of-flight across the cells, column 4

(c) Track a particle bunch.

Use MCOBJET[KOBJ.=3] here, to populate an initial Gaussian distribution with
random transverse coordinates. The periodic optical functions needed in MCOB-
JET are taken from the 10 MeV beam matrix in Table 10.11, namely, .αx = α y = 0,
.β x = 0.7268 m, .β x = 4.4988 m. This yields

to substitute to OBJET in Table 10.12.

426 10 FFAG, Scaling

Fig. 10.27 An acceleration from 10 to 200 MeV. Top row: RF phase (left) and relative momentum
(right) as a function of turn number, over an acceleration cycle. Bottom row: vertical excursion
√ (left)
√ (right). Motion damping is given by Eq. 10.22, namely, . Z damping .∝ R/ p, . P
and vertical angle
damping .∝ 1/ Rp, and normalized invariant . pε Z = constant

10.4.2 RACCAM Proton Therapy Spiral Sector FFAG

This series of exercises concerns the 180 MeV spiral sector proton therapy FFAG
design displayed in Fig. 10.10, and its simulation using FFAG-SPI. The design param-
eters of the ring and of its cell dipole are given in Table 10.2 [22, 26–28]. The cell
geometry is sketched in Fig. 10.28, orbits through a pair of cells are sketched in
Fig. 10.29 as an illustration. Note the presence of field clamps on both sides of the
dipole, these can be simulated in FFAG-SPI, by adding narrow, negative field spiral
sectors adjacent to the main dipole [27].

10.6 Field in a Spiral Sector Dipole

(a, b) Generating a 2D field map. Using TOSCA.
The mid-plane magnetic field can be generated from Eq. 10.15, with geometrical
and field data taken from Table 10.2. This is the FFAG-SPI field model, used here
to produce the field along trajectories. It is based on the field modeling technique
described in Sect. 10.2.2. The input data file is given and commented in Table 10.15,
which can be referred to for details.

The mid-plane field data . B Z (R, θ )|Z=0 are arranged under the form of a 2D even
meshing, this is in order to allow possible handling by TOSCA or POLARMES
10.4 Solutions of Exercises of This Chapter: FFAG, Scaling 427

Fig. 10.28 A sketch of

RACCAM spiral sector
dipole and .2π/10 cell. O is
the center of the ring and the
EFBs form a sector angle . A.
Note that the reference orbit
is not strictly circular, the
bending radius is not
constant along the trajectory
over the .2π/N arc (a line of
constant field is an R-radius
arc, centered on O). Field
clamps can be seen
represented (dashed lines),
on both sides of the dipole

Fig. 10.29 A simulation of a

pair of cells of RACCAM
FFAG ring in zgoubi,
including a few orbits, using
the keyword FFAG-SPI. The
simulation includes field
clamps on both sides of the
dipoles

keywords. In order to generate a field map, from particle raytracing, a set of 29 tra-
jectories is launched, with initial coordinates .Y0 , T0 , Z 0 , P0 and relative momentum
. D = p/ pref defined using OBJET[KOBJ.=1]: they all have initial incidence . T0 = 0,
normal to the . AT = 45.83662◦ angular sector which contains the magnet, whereas
initial radii .Y0 are evenly spaced over the useful field region, namely

Y : r1 → r29 , step /\r

. 0 with r1 = 282 cm, r29 = 340 cm, /\r = 1 cm

Axial motion is taken null, . Z 0 = 0 and . P0 = 0. For each particle, the motion
.

is forced to maintain constant radius, .r ∈ {r1 , r29 }, throughout the dipole, using
OPTIONS[CONSTY ON]. The integration step size is ./\s = 3.46 cm, resulting in
81 steps over

θ : θ1 → θ81 , step /\θ = /\s/R0 ,

. with θ1 = 0, θ81 = AT, /\θ = 0.572906◦
428 10 FFAG, Scaling

Table 10.15 Simulation input data file RACCAMCell.inc: RACCAM .36◦ cell, comprised of a
spiral sector dipole. The FFAG-SPI keyword allows defining up to 5 independent dipoles in an AT
angular sector; only one is defined in the present case, in order to generate field in a single dipole
(note: as FFAG-SPI can house 5 dipoles, field clamps could be simulated by adding a reversed-field
narrow sector on each side of the main dipole). OPTIONS[CONSTY ON] allows to generate a
field map by raytracing a set of trajectories with constant radius. The present file also defines the
sequence segment #S_RACCAMCell to #E_RACCAMCell, for use in INCLUDEs in subsequent
exercises

with radius . R0 = 346.031 cm, and . AT = 45.83662◦ the sector opening. AT includes
.
extra extent, beyond the .36◦ angular extent of a period, in order to avoid cutting off
field tails. In doing so, the angles TE and TS of FFAG-SPI’s KPOS parameters, are
used to re-establish the .36◦ periodicity. This generates the mid-plane field . B Z (R, θ )
over a . Nr × Nθ = 29 × 81 node 2D meshing, as displayed in Fig. 10.30. Note that
if a 3D map is desired instead, a . Z 0 sampling can be added in OBJET, as CONSTY
also forces . Z to its initial value . Z 0 .
10.4 Solutions of Exercises of This Chapter: FFAG, Scaling 429

Fig. 10.30 Using FFAG-SPI theoretical modeling: mid-plane magnetic field in RACCAM spiral
sector dipole, in the laboratory frame (. X = R cos θ and .Y = R sin θ ). The meshing geometry is
obtained by ray-tracing 29 particles forced on circular trajectories evenly spaced in radius with
constant angular integration step size. FFAG-SPI uses the spiral sector analytical field model of
Eq. 10.15

Using TOSCA.
Formatting the field map for TOSCA[. I Z > 1, M O D = 25], and raytracing using
the latter, is left to the reader.1 A dedicated table in [33, TOSCA] explains the choice
[. I Z > 1, M O D = 25], and other options available in the case of a polar field map
mesh. See also Exercises 3.1, 4.1 for guidance.
A similar problem is treated, its input data file and field map are provided, in
zgoubi sourceforge repository, at

https://sourceforge.net/p/zgoubi/code/HEAD/tree/branches/exemples/FFAG/
KEK150MeV/OPERAMapModel
(c) Moving the model to CYCLOTRON.
This exercise is left to the reader. Refer to Exercise 4.6 and to Zgoubi User’s
Guide [33, cf. CYCLOTRON] to work out this simulation.
A CYCLOTRON simulation can be found, input data file is available, in zgoubi
sourceforge repository, at

https://sourceforge.net/p/zgoubi/code/HEAD/tree/branches/exemples/cyclotron/
PSI/usingCYCLOTRON.
10.7 Orbits, Scalloping
(a) Periodic orbits.
The input data file in Table 10.16 will produce 10 closed orbits (found one by one by
FIT) for as many different momenta (REBELOTE repeats the sequence) ranging in

1Note that, as a guidance, TOSCA simulations of all sorts, with various types of data formatting—
IZ, MOD and MOD2 options—can be found in zgoubi sourceforge repository, https://sourceforge.
net/p/zgoubi/code/HEAD/tree/branches/exemples/KEYWORDS/TOSCA/ folder.
430 10 FFAG, Scaling

Table 10.16 Simulation input data file: find the closed orbit for a set of different momenta.
The INCLUDE is the FFAG-SPI spiral dipole segment from Table 10.15, within LABEL1s
.#S_R ACC AMCell and .#E_R ACC AMCell. Once REBELOTE do-loop is completed, SYSTEM
launches a subsequent zgoubi job, plotOrbits.dat, in an ad hoc temporary folder, ./tempo
10.4 Solutions of Exercises of This Chapter: FFAG, Scaling 431

Fig. 10.31 Periodic orbits and field across RACCAM FFAG spiral sector dipole. Left: orbit scal-
loping across AT.=45.83.◦ arc extent in FFAG-SPI simulation (a cell is 36.◦ ), for different proton
energies ranging in 15–180 MeV. Right: field experienced along these orbits, increasing with energy

(relative to . pref = 608.422 MeV/c, 180 MeV proton) . p/ pref = 0.2768526 : 1 (15–
180 MeV). For each closed orbit, coordinates are stored in orbits.fai file (at MARKER
with LABEL1.=afterFIT right after the FIT, prior to REBELOTE repeat).
The input data file ends with a SYSTEM command which, once zgoubi is done
with finding/storing the periodic orbits, launches a subsequent computation (“cd
tempo; ./zgoubi -in plotOrbits.dat” command) which performs the following:

– first, the periodic orbit coordinates are read from orbits.fai,

– they are then pushed through the magnet, and the trajectories are logged in
zgoubi.plt (the effect of IL.=2 under FFAG) for further post-treatment or plot-
ting.

A plot is launched by the next two gnuplot commands under SYSTEM, outcomes
are displayed in Fig. 10.31.
A plot of orbits around the ring can be obtained from the previous raytracing, for
instance using a loop in gnuplot to increment the polar angle by steps of .36◦ , reading
particle data across the cell from zgoubi.plt; it is displayed in Fig. 10.32.

(b) Homothety-rotation of the orbits.

The orbit scalloping is apparent in Figs. 10.31 and 10.32. Step By Step values can
be drawn from zgoubi.plt and show that the scalloping .δ R/R is in the % range.
Rotation of the closed orbit pattern is also apparent in Figs. 10.31 and 10.32; the
radius dependence of the rotation angle satisfies Eq. 10.14. The expected value from
the latter can be checked against closed orbit data from zgoubi.plt.
(c) Figure 10.33 compares the numerical and theoretical (Eq. 10.8) values of the
average orbit radius . R = C/2π , both in good accord.
432 10 FFAG, Scaling

1
YLab [m]

0
0 1 2 3 4

-1

-2

-3

-4
-4 -3 -2 -1 0 1 2 3 4
XLab [m]

Fig. 10.32 Fifteen closed orbits around the 10-cell spiral sector RACCAM ring, in the range 15 to
180 MeV

Fig. 10.33 Dependence of the average closed orbit radius . R = C/2π on the relative momentum.
Markers are from one-turn raytracing. The solid line is from theory, for comparison, after Eq. 10.8
taken for .k = 5 (Table 10.2) and reference momentum . pref = 608.422 MeV/c (180 MeV, radius
. R0 = 3.46031 m). They differ by .≈ 1%
10.4 Solutions of Exercises of This Chapter: FFAG, Scaling 433

10.8 Zero-Chromaticity
(a) Momentum dependence of tunes.
The input data file in Table 10.17 computes momentum-dependent transport matrices
of the cell, .[Ti j ]( p), for a series of different momenta.
OBJET[KOBJ.=5] defines a set of 13 particles with proper initial coordinates
for matrix computation, by MATRIX. Prior to matrix computation, the momentum-
dependent closed orbit is found by FIT. REBELOTE changes the relative momentum
D in OBJET, and repeats the procedure. MATRIX[PRINT] logs the transport coef-
ficients to zgoubi.MATRIX.out, together with the beam matrix and tunes which are
obtained from the hypothesis of periodicity (specified via MATRIX[IFOC.=10+1
period]), namely from the identification (Sect. 14.5.2)

.[Ti j ] = I cos μ + J sin μ

Table 10.17 Simulation input data file: compute the first order transport matrix of the cell for a series
of momenta. Prior to matrix computation, the closed orbit is found by FIT, or FIT2. The INCLUDE
is the FFAG-SPI spiral dipole segment from Table 10.15, within LABEL1s .#S_R ACC AMCell
and .#E_R ACC AMCell
434 10 FFAG, Scaling

Fig. 10.34 Radial (.ν R ) and axial (.ν Z ) tunes of the 10-cell ring (left vertical scale), as a function of
relative momentum (. p/ pref = 1 for 180 MeV). The penalty values (scattered squares, right vertical
scale) monitor the FIT runs, they have to be small to confirm the convergence of FIT. Left: case
( )−0.52
of slowly increasing gap height (thus increasing fringe extent) with radius: .g(R) = g0 RR0 .
Right: case(of linear
) increase of gap height (thus linear increasing of fringe extent) with radius:
. g(R) = g0
R0 , .ν Z is now constant
R

These data are then read and plotted (Fig. 10.34). The radial tune .ν R is constant (apart
from a slight depression towards lower momenta where the dipole width reduces
due to the spiral shape) as expected from the zero-chromaticity resulting from the
scaling law (momentum-independent index, Eq. 10.3). Such is not the case for the
axial tune, .ν Z .
(b) . R-dependence of axial tune.
The variation of the axial tune with momentum, as observed in (a), is due to the fact
that the fringe field extent does not increase fast enough with radius. Indeed the gap
shape index in the present spiral sector model is .κ = −0.52 (cf. simulation input
data file in Table 10.15), resulting, in the FFAG-SPI modeling, in a gap height
( )−0.52
R0
. g(R) = g0
R

whereas constant wedge angle focusing requires the fringe field extent to be propor-
tional to . R, i.e. .κ = −1, yielding
( )
R
. g(R) = g0
R0
10.4 Solutions of Exercises of This Chapter: FFAG, Scaling 435

(c) Constant axial tune.

The axial tune is constant if the gap height is proportional to R [22, 26]. The simu-
lation is obtained by changing the lines concerned in FFAG-SPI keyword input data
list (Table 10.15), namely

change “8.95 − 0.52 ! EFB [etc.]”

. to “8.95 − 1. ! EFB [etc.]”

at the 2 EFBs. The resulting tunes, for this .g(R) ∝ R gap shape, are displayed in
Fig. 10.34.
(d) Momentum compaction and transition .γtr
TWISS keyword is used concurrently with OBJET[KOBJ.=5], to compute vari-
ous first and second order optical parameters including chromaticities. MATRIX
[IORD.=2, IFOC.=11] concurrently with OBJET[KOBJ.=6] can be used as well. A
typical input file is given in Table 10.18. Results are given in Table 10.19.
The momentum compaction is expected to satisfy .α = /\C C
/ /\p
p
= 1/(1 + k). In
the present design .k = 5, yielding .α = 0.1666. This approximation is acceptable at
high energy where computed .α = 0.1666 (Table 10.19), fairly close to .1/(1 + k),
however it is not at low energy where the numerical integration yields .α = 0.60.
This discrepancy might result from the greater flutter at smaller orbit radius (shorter
magnet body compared to fringe field extent)—further investigation is left to the
reader.

Table 10.18 Simulation input data file: TWISS command, to obtain the periodic beam matrix,
momentum compaction, chromaticities, etc. The initial reference coordinates, under OBJET, are
for 15 MeV. TWISS proceeds in 3 stages: it first computes tunes of an on-momentum particle,
then for .±δp/ p off-momentum particles; at each stage, FIT ensures that the reference particle (1st
particle of the 11-set) is on the closed orbit. The INCLUDE uses the segment #S_RACCAMCell
to #E_RACCAMCell defined in Table 10.15
436 10 FFAG, Scaling

Table 10.19 Outcomes of TWISS computation out of zgoubi.res execution listing, including beam
matrix, tunes, momentum compaction factor, and near-zero chromaticities

10.9 Beam Envelopes, Optical Functions

Beam envelopes can be obtained by single particle raytracing over a few tens of
passes through a cell. A proper data file for that is given in Table 10.20, it uses
OBJET[KOBJ.=2] to define the particle.
Multiturn raytracing results are given in Fig. 10.35, which also displays the square
of the transverse particle excursion, normalized to the motion invariant, to yield
betatron function amplitudes, namely, .βY = Y 2 /εY /π and .β Z = Z 2 /ε Z /π .
10.4 Solutions of Exercises of This Chapter: FFAG, Scaling 437

Table 10.20 Simulation input data file: raytrace two particles with different momenta through a
cell, over 110 passes. Initial radial and axial coordinates are taken on .εY,Z /π = 10−8 m invariants.
The INCLUDE uses the segment [#S_RACCAMCell:#E_RACCAMCell] defined in Table 10.15

Fig. 10.35 Left: axial excursion. Z (s) of two particles, one 15 MeV and the other 180 MeV, over 110
passes across a 45.83.◦ arc (the extent of the field region, AT, in FFAG-SPI, whereas a period is 36.◦ ).
Both particles are taken on an axial invariant .ε Z /π = 10−8 m. This multiple pass plot generates
the beam envelopes: the extrema of particle excursion. Right: a graph of . Z (s)2 /ε Z , whose extreme
value represents the betatron function amplitude .β Z (s). Graphs obtained using zpop, left: menu
7; 1/1 to open zgoubi.plt; 2/[8,4] to select . Z versus . X (azimuthal angle); 7 to plot; right: menu 7:
3/14 to change the axial coordinate to . Z 2 /constant

An alternate technique to get the optical functions at all .s across the cell is by
transporting the beam matrix from the origin (.s0 )

σ (s) = T (s ← s0 ) σ (s0 ) T̃ (s ← s0 )
.

The transport matrix .T (s ← s0 ) can be computed step by step from the particle
coordinates stored in zgoubi.plt during the raytracing. A tool in zgoubi toolbox just
does that, betaFromPlt [36], it requires using the 13-particle OBJET[KOBJ.=5], and
logging stepwise particle data in zgoubi.plt, using FFAG-SPI[IL.=2] or equivalently
OPTIONS[.plt 2]. This method is used in various other exercises, which can be
referred to.
438 10 FFAG, Scaling

Fig. 10.36 A scan of .k and .ζ . Left: .(k, ζ ) stability domain, right: corresponding .(ν R , ν Z ) stability
domain. In both diagrams a particular working point, .(k, ζ ) = (4.415, 50.36), is shown for illus-
tration (different from the working point in these exercises, which is .(k, ζ ) = (5, 53.7)) A matrix
code was also used for this scan, results differ noticeably, they are displayed here for comparison

10.10 Periodic Stability Domain

The stability domain in the tune diagram, and the corresponding .(k, ζ ) domain, are
obtained by a .(k, ζ ) scan, performed for some arbitrary momentum, for instance
half-way between injection and top energy. A similar simulation input data file to
that in Table 10.17 is used. The process is the following:
(1) a FIT and MATRIX sequence finds closed orbit and related tunes, for a given
.(k, ζ ) value,
(2) REBELOTE then varies k and repeats the FIT & MATRIX sequence (1),
(3) from that scan results a series of .(ν R , ν Z ) couples, corresponding to a series of
.(k, ζ ) couples at fixed spiral angle .ζ .

This (1)–(3) sequence is repeated for a series of .ζ values—an external program can
be used to perform that iteration on .ζ .
This results in .(k, ζ ) and .(ν R , ν Z ) stability diagrams displayed in Fig. 10.36. The
correlation comes out to be, mostly, an increase of .ν R with .k and increase of .ν Z with
.ζ , however the two quantities are not fully decoupled, increasing .k (respectively, .ζ )
has a slight effect on .ν Z (resp. .ν R ).
10.11 Motion Stability Limit
The input data file in Table 10.21 can be used:
– raytracing is performed for one particle at a time (namely, for a particular energy
taken in [15 MeV, 180 MeV]),
– REBELOTE performs a multiturn raytracing.
Then push the initial coordinate (.Y0 for radial stability limit, . Z 0 for axial), up to
stability limit. An external program available in zgoubi toolbox, searchStabLim,
can be used for that [37].
This will result in the horizontal and vertical phase space portraits displayed in
Fig. 10.12.
10.4 Solutions of Exercises of This Chapter: FFAG, Scaling 439

Table 10.21 Simulation input data file: track one particle, for 1000 turns. Push .Y0 up to find the
stability limit. Using the complete ring (10 cells) allows introducing non-systematic errors if desired
(random field or positioning errors for instance, using ERRORS keyword). If only systematic errors
or non-linearities are of interest, then a single cell is used

10.12 Dynamic Aperture Scan

A DA scan is obtained by repeating the previous (Exercise 10.11) stability limit
search:

– first, look for the maximal radial extent .[xmin , xmax ] of stable horizontal motion,
at quasi-zero axial invariant,
– second, look for the maximum stable axial amplitude, at various values .x ∈
[xmin , xmax ].

An external program available in zgoubi toolbox does that, searchDA [38]. The
exercise results in the DA graph of Fig. 10.37.

Fig. 10.37 Dynamic aperture in the (Y,Z) space, at 15, 57 and 180 MeV. The origin here, x .= 0, is
on the closed orbit at the momentum of concern
440 10 FFAG, Scaling

10.13 Acceleration, Transverse Betatron Damping

A similar problem is solved for the radial FFAG, it can be referred to (Exercise 10.5),
in addition to the guidance below.
Setting up the acceleration requires the following:

– PARTICUL[PROTON] is necessary as CAVITE is used: it allows converting

energy change in rigidity change (zgoubi pushes particles using rigidity),
– SCALING[IOPT.=-2] provides the RF program to CAVITE (by reading it from
an external file, zgoubi.freqLaw.In),
– CAVITE[IOPT.=6] boosts the particle(s) at each pass, following that pre-defined
RF program,
– REBELOTE do-loop sends the execution pointer back to the top of the input data
file, for multiturn tracking. A 10 kV acceleration rate per turn may be obtained from
constant peak voltage .V̂ = 20 kV and synchronous phase .φs = 30◦ , this deter-
mines the number of passes for a 15.→180 MeV cycle, namely,.(180 − 15)/0.01 =
16500, hence a number of repeat REBELOTE[NPASS.=16499],
– FAISTORE[FNAME.=zgoubi.fai, IP.=1] stores turn-by-turn particle data in
zgoubi.fai, every turn.

For simplicity the RF program may be limited to the turn dependence of RF

frequency (peak voltage and synchronous phase maintained constant). Elaborat-
ing zgoubi.freqLaw.In (essentially, here, determining revolution period versus turn
number) requires the following steps (Exercise 10.5 may also be referred to, see
zgoubi.freqLaw.In file formatting and sample content in Table 10.14):
(i) run a search for closed orbits for a few tens of energies in the range 15 to
180 MeV. Time of flight, derived from path length and particle velocity, is part
of the outcome of orbit computation as PARTICUL provides necessary particle
data in that aim. That yields the turn dependence of revolution period, namely,
. f RF = h × f rev = v/C (assuming RF harmonic .h = 1). Note that in the absence
of orbit defects or other tailored bumps, the time of flight may be obtained with
1
good accuracy from the theoretical orbit length .C( p) = C0 ( p/ p0 ) k+1 (Eq. 10.8);
(ii) re-write these turn-time of flight data, properly formatted (cf. Table 10.14), in
zgoubi.freqLaw.In. Note that zgoubi.freqLaw.In does not need to contain all
turns, zgoubi will interpolate.
The resulting radial and axial phase spaces are displayed in Fig. 10.38. The beta-
tron damping satisfies Eqs. 10.22, 10.23. The homothety-rotation of the geometrical
scaling is apparent in the axial phase space portrait.
References 441

Fig. 10.38 Acceleration from 15 to 180 MeV. Radial and axial phase spaces. Graphs obtained using
zpop: menu 7; 1/5 to open zgoubi.fai; 2/[2, 3] to select .T (angle) versus .Y (radius) [or 2/[4, 5], for
. P versus . Z ]; 7 to plot

10.4.3 FFAG Acceleration Methods

In the following three exercises, solutions are based on input data files worked out in
Sect. 10.3.1 exercise series, with minor adaptations. Regarding beam acceleration,
input data files developed as part of Exercises 10.5, 10.13 are used.

10.14 Hybrid Acceleration

Refer to [29, 30] where all necessary details regarding working hypotheses, and
partial results, including numerical simulations, can be found.
Exercise 11.1-c may also be used as a guidance, it simulates serpentine accelera-
tion in the linear FFAG ring EMMA.

10.15 Bucket Acceleration

Refer to [39] where all necessary details regarding working hypotheses, and partial
results, including numerical simulations, can be found.
Regarding in-flight decay method and outcomes, refer to the exercises in
Sects. 12.3.3, 12.3.4.

10.16 Serpentine Acceleration

Refer to [40, 41] where all necessary details regarding working hypotheses, and
partial results, including numerical simulations, can be found.

References

1. F.T. Cole, O Camelot, a Memoir of the MURA Years. Cyclotron Conference, East Lansing,
USA, May 13–17, 2001. https://accelconf.web.cern.ch/accelconf/c01/cyc2001/extra/Cole.pdf
2. A.A. Kolomensky, et al., Some questions of the theory of cyclic accelerators. Ed. AN SSR,
1955, p. 7, PTE, No 2, 26 (1956)
442 10 FFAG, Scaling

3. K.R. Symon et al., Fixed-field alternating-gradient particle accelerators. Phys. Rev. 103, 1837
(1956)
4. T. Ohkawa, Two-beam fixed field alternating gradient accelerator. Rev. Sci. Instrum. 29, 108
(1958); A concept presented at a meeting of the Physical Society of Japan in 1953 (1967)
5. K.R. Symon, MURA days, in Proceedings of the 2003 Particle Accelerator Conference. https://
accelconf.web.cern.ch/p03/PAPERS/WOPA003.PDF Figs. 10.1 and 10.9: Copyrights under
license CC-BY-3.0, https://creativecommons.org/licenses/by/3.0; no change to the material
6. L. Jones, F. Mills, A. Sessler, K. Symon, D. Young, Innovation Was Not Enough; A History of
the Midwestern University Research Association (MURA) (World Scientific, 2010)
7. Y. Mori, Developments of FFAG accelerator, in 17th International Conference on Cyclotrons
and Their Applications 2004, Tokyo (Japan), 18–22 October 2004. https://www.osti.gov/
etdeweb/biblio/20676358
8. M. Craddock, The rebirth of the FFAG. CERN Courrier (27 July 2004). https://cerncourier.
com/a/the-rebirth-of-the-ffag/
9. J. Collot, The rise of the FFAG. CERN Courrier (19 August 2008). https://cerncourier.com/a/
the-rise-of-the-ffag/
10. S. Machida, Muon (FFAG) accelerators. THYAB01 talk; PAC 2007 Accelerator Confer-
ence, June 25–29, 2007, Albuquerque, NM, USA. https://accelconf.web.cern.ch/p07/TALKS/
THYAB01_TALK.PDF. Figures 10.2, 10.6: Copyrights under license CC-BY-3.0, https://
creativecommons.org/licenses/by/3.0; no change to the material
11. Y. Ishi, Status of KURRI facility, in Proceedings of the FFAG 2016 Workshop, Impe-
rial College, London (2016). https://indico.cern.ch/event/543264/contributions/2295846/
attachments/1333675/2005286/FFAG16_LONDON_ishi.pdf
12. C.H. Pyeon et al., First injection of spallation neutrons generated by high-energy protons into
the Kyoto University critical assembly. J. Nucl. Sci. Technol. 46, 1091 (2009)
13. F. Méot, A multiple-room, continuous beam delivery, hadron-therapy installation.
Phys. Procedia 66, 361–369 (2015). https://www.sciencedirect.com/science/article/pii/
S1875389215001984
14. E. Yamakawa, et al., High intensity proton FFAG ring with serpentine acceleration for ADS,
in MOP209 Proceedings of HB2012, Beijing, China. https://accelconf.web.cern.ch/HB2012/
papers/mop209.pdf
15. E. Yamakawa et al., Serpentine acceleration in zero-chromatic FFAG accelerators. Nucl.
Instrum. Methods Phys. Res., Sect. A 716(11), 46–53 (2013). (July)
16. T. Planche, et al., New approaches to Muon acceleration with zero-chromatic FFAGs,
in THPD093, Proceedings of IPAC’10, Kyoto, Japan (2010). http://accelconf.web.cern.ch/
AccelConf/IPAC10/papers/thpd093.pdf
17. Y. Mori, et al., Multi-beam acceleration in FFAG synchrotron, in Proceedings of the PAC
2001 Accelerator Conference, pp. 588–590 (2001). http://accelconf.web.cern.ch/AccelConf/
p01/PAPERS/ROPA010.PDF
18. A. Sato, et al., FFAG as phase rotator for the PRISM project, in Proceedings of the EPAC
2004 Accelerator Conference, pp. 713–715 (2004). http://accelconf.web.cern.ch/AccelConf/
e04/PAPERS/MOPLT070.PDF
19. M. Haj Tahar, F. Méot, Tune compensation in nearly scaling fixed field alternating gradi-
ent accelerators. Phys. Rev. Accel. Beams 23, 054003 (2020). https://journals.aps.org/prab/
abstract/10.1103/PhysRevAccelBeams.23.054003
20. K. Okabe, et al., Development of H-injection of proton-FFAG at Kurri, in THPEB009 Pro-
ceedings of IPAC’10, Kyoto, Japan. https://accelconf.web.cern.ch/IPAC10/papers/thpeb009.
pdf Fig. 5.1: Copyrights under license CC-BY-3.0, https://creativecommons.org/licenses/by/
3.0; the photo has been trimmed to mostly leave the 2.5 MeV injector
21. D. Neuvéglise, F. Méot, An alternative design for the RACCAM magnet with distributed
conductor, in FR5REP095, Proceedings of the PAC09 Conference, pp. 5002–5004, Vancouver,
BC, Canada (2009). https://accelconf.web.cern.ch/PAC2009/papers/fr5rep095.pdf
22. T. Planche et al., Design of a prototype gap shaping spiral dipole for a variable energy proton
therapy FFAG. NIMA 604, 435–442 (2009)
References 443

23. M. Aiba, et al., Development of 150-MeV FFAG at KEK, in FFAG03 Workshop, KEK, Japan,
July 7–12, 2003. https://ffag.pp.rl.ac.uk/FFAG/FFAG03_HP/index.html
24. M. Aiba, F. Méot, Determination of KEK 150 MeV FFAG parameters from ray-tracing
in TOSCA field maps, CERN-NUFACT-NOTE-140; CARE-Note-2004-030-BENE; CEA-
DAPNIA-2004-188-2004. http://cds.cern.ch/record/806545/files/note-2004-030-BENE.pdf
25. M. Aiba, Tracking study for FFAG, in FFAG Accelerator Workshop; FFAG02, KEK, Tsukuba
February 13–15, 2002. https://ffag.pp.rl.ac.uk/FFAG/FFAG02_HP/2002_02_13/20020213_
M.Aiba.pdf
26. S. Antoine et al., Principle design of a proton therapy, rapid-cycling, variable energy spiral
FFAG. NIM A 602, 293–305 (2009)
27. J. Fourrier, F. Martinache, F. Méot, J. Pasternak, Spiral FFAG lattice design tools, application
to 6-D tracking in a proton-therapy class lattice. NIM A 589, 133–142 (2008)
28. F. Méot, RACCAM: a status including magnet prototyping and magnetic measurements, in
International conference on FFAGs, Fermilab, 21–25 September 2009. https://indico.fnal.gov/
event/2672/contributions/77834/attachments/48652/58457/FMeot1-090921.pdf
29. H. Tanaka, Feasibility study of hybrid accelerator and superconducting FFAG, in FFAG04
Accelerator Workshop, KEK, Tsukuba (October 13–16, 2004). http://130.246.92.181/FFAG/
FFAG04_HP/index.html
30. H. Tanaka, et al., Hybrid accelerator using an FFAG injection scheme, in Cyclotrons 2004
Conference, Tokyo, Japan (October 18–22, 2004). http://accelconf.web.cern.ch/AccelConf/
c04/data/CYC2004_papers/19C6.pdf
31. M. Haj Tahar, High Power Ring Methods and Accelerator Driven Subcritical Reactor Appli-
cation. Ph.D. thesis dissertation, BNL and University Grenoble-Alpes (January 2017). https://
www.bnl.gov/isd/documents/94721.pdf. https://www.osti.gov/biblio/1351800
32. E. Yamakawa, et al., Serpentine acceleration in scaling FFAG, in Proceedings of FFAG12 work-
shop, Osaka, 2012. https://indico.cern.ch/event/194713/contributions/1473080/attachments/
282693/395230/FFAG12Slides.pdf Additional details in: https://repository.kulib.kyoto-u.ac.
jp/dspace/bitstream/2433/174929/2/D_Yamakawa_Emi.pdf
33. F. Méot, Zgoubi Users’ Guide. https://www.osti.gov/biblio/1062013-zgoubi-users-guide. An
up-to-date version of the guide can be found at: https://sourceforge.net/p/zgoubi/code/HEAD/
tree/trunk/guide/Zgoubi.pdf
34. F. Lemuet, Collection and Muon Acceleration in the Neutrino Factory Project. Ph.D. The-
sis dissertation, CEA and CERN, Paris-Saclay University, April 2007. https://inis.iaea.org/
collection/NCLCollectionStore/_Public/42/013/42013892.pdf
35. F. Méot, F. Lemuet, Developments in the ray-tracing code Zgoubi for 6-D multiturn tracking
in FFAG rings. NIM A 547, 638–651 (2005)
36. From Zgoubi toolbox, part of the sourceforge package: a Fortran tool to compute optical
functions from a zgoubi.plt output file, and some related gnuplot scripts: https://sourceforge.
net/p/zgoubi/code/HEAD/tree/trunk/toolbox/betaFromPlt/
37. From Zgoubi toolbox, part of the sourceforge package: a Fortran tool to perform a dynamic
aperture scan, and some related gnuplot scripts: https://sourceforge.net/p/zgoubi/code/HEAD/
tree/trunk/toolbox/searchStabLim/
38. From Zgoubi toolbox, part of the sourceforge package: a Fortran tool to perform a dynamic
aperture scan, and some related gnuplot scripts: https://sourceforge.net/p/zgoubi/code/HEAD/
tree/trunk/toolbox/searchDA/
39. T. Planche, et al., New approaches to muon acceleration with zero-chromatic FFAGs,
in THPD093, Proceedings of IPAC’10, Kyoto, Japan (2010). http://accelconf.web.cern.ch/
AccelConf/IPAC10/papers/thpd093.pdf
40. E. Yamakawa, Serpentine acceleration in zero-chromatic FFAG with long straight section, in
International Workshop on FFAG Accelerator (FFAG’10), Kyoto University Research Reac-
tor Institute, Osaka, Japan (28–31 October 2010). http://130.246.92.181/FFAG/FFAG10_HP/
slides/Sat/Sat14Yamakawa.pdf
41. E. Yamakawa et al., Serpentine acceleration in zero-chromatic FFAG accelerators. Nucl.
Instrum. Methods Phys. Res., Sect. A 716(11), 46–53 (2013). (July)
444 10 FFAG, Scaling

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0
International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing,
adaptation, distribution and reproduction in any medium or format, as long as you give appropriate
credit to the original author(s) and the source, provide a link to the Creative Commons license and
indicate if changes were made.
The images or other third party material in this chapter are included in the chapter’s Creative
Commons license, unless indicated otherwise in a credit line to the material. If material is not
included in the chapter’s Creative Commons license and your intended use is not permitted by
statutory regulation or exceeds the permitted use, you will need to obtain permission directly from
the copyright holder.
Chapter 11
FFAG, Linear

Abstract This chapter is an introduction to linear Fixed-Field Alternating Gra-

dient (FFAG) cyclic accelerators. It begins with a brief reminder of the historical
and technological context, and continues with the theoretical material needed for
the simulation exercises. It relies on charged particle optics and acceleration con-
cepts introduced in the previous synchrotron and scaling FFAG chapters and further
addresses

– design aspects of linear FFAGs,

– diverse specificities of linear FFAG optics,
– beam dynamics in rings,
– serpentine acceleration.
Simulations do not require specific keywords, they use optical elements met in the
previous chapters, including MULTIPOL DIPOLE[S], CAVITE, data input/output
keywords such as FAISCEAU, FAISTORE, the SYSTEM keyword, etc. Beam
dynamics simulations include
– computation of optical parameters,
– particle trajectories through a cell or a ring,
– closed-orbit finding, from multi-turn raytracing or by coordinate matching,
– deriving ancillary outcomes from rays, such as
• transport matrices using MATRIX,
.

• periodic optical functions and their transport using TWISS,

.

– finding dynamical aperture,

– serpentine fast acceleration, and more.

© The Author(s) 2024 445

F. Méot, Understanding the Physics of Particle Accelerators, Particle Acceleration
and Detection, https://doi.org/10.1007/978-3-031-59979-8_11
446 11 FFAG, Linear

Notations Used in the Text

.a gyromagnetic anomaly of electron, .a = 1.15965 × 10−3

.B; . Bx,y,s ; . B0 field vector; components in moving frame; reference . R0
. Bρ; Bρ0 particle rigidity: . Bρ = p/q; for reference momentum . p0
.C; .C0 orbit length: .C = ds = 2π R; for reference momentum . p0
.E particle energy
. f rev , . f rf revolution and accelerating voltage frequencies
.H longitudinal Hamiltonian
.h harmonic number, an integer, .h = f rf / f rev
.m; .m 0 ; M particle mass; rest mass; mass in .eV /c2 units
.N number of cells in a ring
. p; p0 ; δp; ∆p particle momentum; reference momentum; offset
.q particle charge
.R average orbit radius: . R = C / 2π
.R F Radio-Frequency
.s path variable
.v particle velocity
. V ; . V̂ accelerating voltage; peak value [ ]
x ', y' particle coordinates in the moving frame . (∗)' = d(∗)/ds
.α momentum compaction, or trajectory deviation
.β = v/c; .β0 ; .βs normalized velocity; reference; synchronous
.βu , αu , γu , D x optical functions (.u : x, y, l), dispersion
.γ Lorentz relativistic factor: .γ = E/m 0 c2 = E[eV ]/M
.δ; .δp/ p momentum offset; relative
.εu Courant-Snyder invariant; beam emittance (.u : x, y, l)
.η; ηi phase-slip factor; coefficients of polynomial in .δ (i .= 0, 1,
2, …)
.θ azimuthal angle
ν
. u wave numbers, horizontal, vertical, synchrotron (.u : x, y, l)
.ωrf angular RF, .ωrf = 2π f rf
.ρ curvature radius
.φ RF phase

11.1 Introduction

The concept of linear FFAG is one of the latest innovations in cyclic accelerators.
It was devised in the late 1990s, in the context of lattice R/D and rapid acceleration
regarding short-lived muon beams for Higgs and other Neutrino Factory.
11.1 Introduction 447

A linear FFAG ring for rapid acceleration leans on

– the use of linear magnets: quadrupoles and/or combined function dipoles

[1, 2] (Fig. 11.1)—by contrast with scaling FFAG optics which uses highly non-
linear AG dipoles, Chap. 10. As in the latter, the magnetic field is fixed during
acceleration;
– quasi-isochronous acceleration using high gradient and fixed RF. A technique
known as serpentine acceleration, or gutter acceleration [3], allowing cyclotron-
style CW operation.

Light-source style chromatic invariant minimization allows minimizing the disper-

sion function and thus dipole size [4].
As a consequence of the use of fixed field linear magnets, the betatron tunes
decrease during acceleration. Furthermore, a large cell tune variation may be sought,
within a half integer range, in order to allow largest momentum sweep over the
acceleration range. A consequence of tune variation during acceleration is that many
horizontal and vertical integer tune values have to be crossed, which essentially limits
the use of this optics to rapid acceleration systems.
Two outcomes of the linear AG method are, a large transverse dynamical accep-
tance intrinsic to a linear lattice, and, as a result of the small dispersion function, a
small transverse extent of magnets in comparison to scaling FFAG technology.

Fig. 11.1 An early ½ F D ½ F fixed-field AG FODO cell design for a 4–16 GeV rapid muon accel-
erator, for injection into a .μ+ − μ− Higgs Factory collider [2]. Case of 4 GeV optics here—optical
functions change with energy as the field is fixed
448 11 FFAG, Linear

Fig. 11.2 A schematic of a

neutrino factory [5]. Proton
bunches in the GeV range hit
a pion production target at a
.∼50 Hz repetition rate. Pions
decay into muons in a few
tens of meters. They are
distributed in a huge 6D
phase space (see Beam Lines
Chap., Figs. 12.17 and
12.18). Bunching and
cooling follow, prior to
200 MeV to 5 GeV
acceleration in linacs, and up
to 10–30 GeV in FFAG1 and
FFAG2. The latter are filled
with high gradient 200 MHz
SCRF cavities

A design of a Neutrino Factory based on the linear FFAG technology for the rapid
acceleration of muon beams was completed in the early 2000s (Fig. 11.2) [5]. An
experimental electron model of a linear FFAG ring was proposed in that context [7],
it was built and commissioned at Daresbury Laboratory in the 2007–2012 period
[8, 9], Fig. 11.3.
Further outcomes of the linear FFAG method include the design of the circular
return arcs of a multiple-pass energy recovery linac, eRHIC, to be located in RHIC
heavy ion collider tunnel, providing 21 GeV electron bunches for an electron-ion
linac-ring collider at the Brookhaven National Laboratory [10–12] (Fig. 11.4). A
prototyping of eRHIC FFAG ERL return loop was proposed and led to the CBETA

Fig. 11.3 Left: EMMA FD cell, and 1.3 GHz RF cavity insert at the center. Right: 42-cell, 16.57 m
circumference EMMA ring [6]. Nineteen 1.3 GHz RF cavities allow 2 MV of acceleration voltage.
They are distributed every other cell (left picture), except for two intervals dedicated to injection
and extraction systems. EMMA was designed for 10–20 MeV electron acceleration in a few turns
11.1 Introduction 449

Fig. 11.4 eRHIC 21.2 GeV ERL project of a linac-ring collider, with its two linear FFAG recircu-
lation loops, respectively 1.3–5.3 GeV (FFAG1) and 6.6–21.2 GeV (FFAG2), alongside RHIC ion
rings. The 1.322 GeV SCRF linac is located in RHIC interaction region 2, it is connected to the
FFAG loops by a merger (respectively spreader) section at its upstream (resp. downstream) end

Fig. 11.5 CBETA ERL [14]. The permanent magnet 8-pass linear FFAG loop recirculates the 42,
78, 114 and 150 MeV beams in a single 2-in diameter vacuum pipe, up and down for energy recovery

ERL experiment [13, 14] (Fig. 11.5). CBETA, comprised of a proof-of-principle

8-pass (4 up, 4 down) permanent magnet return loop [15], was built and commis-
sioned in the 2015–2020 period, it was aimed at demonstrating a factor of 4 in energy,
and full energy recovery.
Studies in the same line are carried on today, regarding application of the linear
FFAG return arc method to a 24 GeV energy upgrade at CEBAF 12 GeV RLA [16].
450 11 FFAG, Linear

11.2 Basic Concepts and Formulæ

A linear FFAG cell is comprised of quadrupoles and/or combined function dipoles,

with alternating field gradients in the range of T/m [1, 2]. This makes it akin to
alternating gradient (AG) strong focusing optical systems. The theoretical material of
transverse periodic stability in beam lines and rings is that of AG focusing (Chap. 9).
Acceleration in rings is based on the quasi-isochronous serpentine method [3],
akin to fixed-frequency, high gradient acceleration in cyclotrons (Chaps. 3 and 4),
yet using higher RF.
In the following sections some specificities of beam optics and acceleration in
linear FFAGs are addressed.

11.2.1 Linear FFAG Cell

A linear FFAG cell is generally comprised of a doublet (FODO, FD) or triplet (FDF,
DFD) of quadrupoles with some radial offset, and/or combined function dipoles. Drift
spaces allow room for instrumentation, RF systems, etc. FFAG cells are designed
to feature a large momentum acceptance, . pmax / pmin up to a factor 4 or more. Any
periodic orbit within the momentum acceptance of the cell experiences a curvature
.2π/N , so that . N cells make up a ring. Depending on momentum, the orbit curvature

may have the same sign, or opposite signs in the cell magnets. Periodic orbits with
increasing momentum move almost everywhere from an inner to an outer radial
excursion. These basic aspects are illustrated in Fig. 11.6 with the case of the FD
cell of the Neutrino Factory 5.→10 GeV FFAG ring [18]. More details regarding the
design of linear FFAG cell magnets, including raytracing in OPERA field maps as
part of optical parameter optimization, can be found in [17].
Two main constraints in designing a cell are, the tune excursion, and quasi-
isochronism of the orbits:

– the natural chromaticity causes orbits of increasing momentum to feature decreas-

ing paraxial tune values (Eqs. 9.18 and 9.19); the rule here is to minimize tune
excursions within a .[0, 0.5] interval (Fig. 11.7);
– in the quasi-isochronous hypothesis the momentum dependence of the time of
flight is tailored to be an about symmetric parabola, centered on the center of the
momentum range (Fig. 11.8), with minimized time excursion so to minimize RF
phase excursion.

Additional design constraints include orbit excursion, field value and gradients,
magnet apertures; fringe fields may require special attention as, often in these struc-
tures, the length/aperture ratio of the magnets may be on the small side. Theoretical
material regarding linear FFAG cell design methods can be found in the Neutrino
Factory technical reports [19] and FFAG workshops [20].
11.2 Basic Concepts and Formulæ 451

Fig. 11.6 Optical properties of the Neutrino Factory 10 GeV FFAG ring [18]. a periodic orbits
across a pair of cells; b their coordinates in phase space at.s = 0; c magnetic field along the periodic
orbits across a cell

Fig. 11.7 Cell tunes

.(νx , ν y ),
decreasing with
increasing beam energy

Fig. 11.8 Quadratic

dependence of time of flight
on energy, in the neutrino
factory .5 → 10 GeV muon
ring
452 11 FFAG, Linear

11.2.2 Quasi-Isochronous Serpentine Acceleration

In cyclotrons, isochronism has to be achieved to a level of typically 0.01%, meaning

a revolution time shifting by 0.01% over injection to extraction energy range. Quasi-
isochronous fixed RF acceleration in an FFAG is based on lattice isochronism at
typically 0.1% level in large rings (Fig. 11.8) or higher in smaller rings (as EMMA,
see Exercise 11.1).
In the quasi-isochronism hypothesis the expression of the phase slip factor
(Eq. 8.33) has to include higher order terms in the momentum deviation [21], namely,

η = η0 + η1 δ + η2 δ 2 + · · ·
. (11.1)

Under these conditions, Eqs. 9.27 take the form

dφ 2π h dδ e V̂
. = (η0 δ + η1 δ 2 + η2 δ 3 + · · · ), = [sin φ − sin φs ] (11.2)
dt Ts dt Ts βs2 E s

From this the Hamiltonian to the second order in .δ can be inferred, namely,

δ2 4 η1 δ 3
H(φ, δ, t) = 2
. + + [cos φ − cos φs + φ sin φs ] (11.3)
∆2 3 η 0 ∆2

where ( )1/2
2e V̂
∆=
. (11.4)
π h|η|βs2 E s

Due to the quadratic dependence of the time of flight (Fig. 11.8), two new longitudinal
phase space fixed points appear, namely (Fig. 11.9),

φfps1 = π − φs , δfps1 = 0 stable fixed points

φfps2 = φs , δfps2 = − ηη01

.
φfpi1 = π − φs , δfpi2 = − ηη01 unstable fixed points

φfpi2 = φs , δfpi1 = 0

This results in two possibilities to accelerate a bunch, namely,

(i) carrying out a half synchrotron oscillation inside a bucket (in the manner of
Fig. 10.14, Sect. 10.2.3);
11.2 Basic Concepts and Formulæ 453

Fig. 11.9 Contour of the

longitudinal gutter motion,
.H(φ, δ) = constant [18],
and the stable (.pfS1,2 ) and
unstable (.pfI1,2 ) fixed points.
Case here of .φs = 0, ∆ = 1,
and .η1 /η0 = 1

(ii) following the channel between buckets, a method called gutter acceleration,
or serpentine acceleration.
The channel width defined by the separatrices, and the energy reach, depend
on .∆ (Eq. 11.4). Figure 11.10 shows longitudinal phase space portraits for various
.∆ values.
Experiments conducted in the EMMA ring (Fig. 11.3) have demonstrated ser-
pentine acceleration in a linear FFAG, and absence of emittance growth upon fast
crossing of integer resonances (nine are crossed on the way from 10 to 20 MeV/c) at a
high acceleration rate (.>1 MV/turn). Difficulties lie in the large chromaticity, which
causes transverse decoherence and thus emittance growth [8], as well as significant
time-of-flight dependence on transverse amplitude which may result in longitudinal
emittance growth [22].

Fig. 11.10 Contour of the longitudinal gutter motion (.φs = 0, η1 /η0 = 1) for, from left to right,
.∆= 0.7, 1 and .1.3
454 11 FFAG, Linear

11.2.3 Synchrotron Radiation

The SR energy loss along an arc has been introduced in Chap. 5, Betatron, Sect. 5.2.3,
as the effect was first observed in that cyclic accelerator. This theoretical material
applies to beam lines, that includes FFAG lines. Beam emittance perturbations are
proportional to the energy spread (Eq. 5.13) .σ E /E ∝ γ 5/2 /ρ, which may result in
substantial emittance growth at high energy.
Assume a bunch traveling along a linear FFAG lattice, for instance an eRHIC
ERL return loop [10], or FFA@CEBAF RLA recirculation arc [16]. Radiation will
occur in the dipole field regions along the trajectory. For simplicity a DF doublet
is considered below, this can easily be extended to a triplet. In the present formal
approach, these magnets can be, indifferently, either offset quadrupoles, or combined
function dipoles.
Over an arc .∆θ with 1/.ρ the curvature, assumed constant, the energy loss
(Eq. 5.12) and energy spread (Eq. 5.13) can be written, respectively [11]
5√
∆E γ 3 ∆θ σE −14 γ
2 ∆θ
. = 1.9 × 10−15 , = 3.8 × 10 (11.5)
E ref ρ E ref ρ

Take for average radius, in the focusing magnet (QF) and in the defocusing magnet
(BD) respectively,
sBD sQF
ρ
. BD ≈ , ρQF ≈
∆θBD ∆θQF

with .sBD and .sQF the arc lengths. Consider in addition, with .lBD , .lQF the magnet
lengths,

s
. BD ≈ lBD , sQF ≈ lQF

This yields, over a cell,

( )
−15 lBD lQF
∆E[MeV ] ≈ 0.96 × 10
. γ 4
+ 2 (11.6)
ρBD
2
ρQF

Taking in addition .< (1/ρ)2 > ≈ 1/ < ρ 2 >, an estimate of the energy spread is
/
−14 lBD lQF
σ ≈ 1.94 × 10
. E γ 7/2
+ 3 (11.7)
|ρBD | |ρQF |
3

The bunch lengthening over a.[s, s f ] distance, resulting from the stochastic energy
loss, can be written [23]
11.2 Basic Concepts and Formulæ 455

Fig. 11.11 Average energy loss (left axis) and energy spread (right axis). Solid lines: theory,
respectively Eqs. 11.6 and 11.7. Markers: . E : 6.322 → 21.164, step 1.322 GeV recirculations,
from tracking with Monte Carlo SR. Lower horizontal scale: .a = 1.16 × 10−3 is the electron
gyromagnetic anomaly

(σ ) [ 1 ʃ sf ( )2
]1/2
E '
.σl = Dx (s)T51 (s f ← s) + Dx (s)T52 (s f ← s) − T56 ds
E L bend s

with the integral taken over the bends, . Dx and . Dx' the dispersion function and its
derivative, .T5i the trajectory lengthening coefficient of the first order mapping (.i =
1, 2, 5, 6 stand for respectively .x, .x ' , .δl, .δp/ p coordinates).
As an illustration, Fig. 11.11 shows the case of the 3.8 km long eRHIC ERL
(energy recovery linac) return loop, comprised of 6 arcs, 120 cells per arc, based
on the cell studied in Exercise 11.2. Eleven beams are circulated in the ring, with
energies . E : 6.322 → 21.164, step 1.322 GeV. The energy dependence of energy
loss shows a local minimum in the .aγ = 30 − 35 region, a different behavior from
the classical .γ 4 dependence in an isomagnetic lattice (Eq. 5.14), due to the large
variation of the curvature radius over the 7.9 .→ 21.2 GeV energy range (Fig. 11.11).

11.2.4 Polarization

Spin dynamics in magnetic fields has been introduced in Sect. 3.2.5. Over long beam
lines and in particular conditions, spin motion may be subject to resonance with
the betatron motion [24] (resonant spin motion is introduced in Sects. 4.2.5, 8.2.4
and 9.2.7). However this should generally not be the case in a beam line and is not
considered here.
Spins of electrons traveling in a vertical guiding field precess by an angle.aγ α (.a =
1.16 × 10−3 is the electron gyromagnetic anomaly) as the velocity vector precesses
by an angle .α (Fig. 3.16). The resulting precession rate in a ring is .aγ per turn
(Eq. 3.33).
456 11 FFAG, Linear

Fig. 11.12 Spin diffusion

out of the bend plane, at each
pass (not cumulated). Case
of a 5,000 electron bunch
tracking (empty circles), and
from Eq. 11.8 (dashed curve)

Considering the aforementioned eRHIC linear FFAG [10], where polarization

orientation is in the bend plane, momentum spread in a bunch is one cause of depo-
larization. Under the effect of radial and longitudinal field components, vertical
motion (orbital or betatron) causes spins to leave the median plane, also causing
depolarization. The stochastic emission of photons induces spin diffusion which
also contributes to depolarizing the beam. The effect is introduced in Sect. 9.2.7. As
far as numerical integration is concerned, spin diffusion is a sub-product of the step-
wise integration of spin motion, accounting for Monte Carlo simulation of photon
emission. The theoretical evolution of the spin diffusion in eRHIC satisfies [25]
⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
∆E 2 1 0 0 ∆E 2 s
. ⎝ ∆E∆φ ⎠ = ⎝ αs 1 0 ⎠ ⎝ ∆E∆φ ⎠ + ω × ⎝ αs 2 /2 ⎠ (11.8)
∆φ 2 α s 2αs 1
2 2
∆φ 2 α 2 s 3 /3
s=0

where .s is the distance in the field, .ω = C λ̄c re γ 5 E 2 /ρ 3 ≈ 1.44 × 10−27 γ 5 E 2 /ρ 3 ,

.α = a /ρ E 0 ≈ 2.27/ρ (with .λ̄c = /m e c the electron Compton wavelength, .C =
√
110 3/144, . E 0 = m e c2 /e the electron rest mass).
⎛ ⎞
∆E 2
Assuming a starting state .⎝ ∆E∆φ ⎠ = 0 (this is the case for each energy
∆φ 2
s=0
1/2 √
for instance in Fig. 11.12) yields .σ E = ∆E 2 = ωs (which in passing identifies
5/2 √
with the familiar .σ E /E = 3.8 × 10−14 γρ 3/2 s), so that
/
1/2 ωα 2 s 3 αs
σ = ∆φ 2
. φ = = √ σE (11.9)
3 3
σφ
or, given s = 2πρ, = 8.23 [rad/GeV/turn]
σE

As an illustration, Fig. 11.12 displays the evolution of the rms value .σφ of the angle
of the spin with respect to the bend plane, in the aforementioned 3.8 km long eRHIC
11.3 Exercises 457

return loop. .σφ is calculated at each pass, for the 12 different energies . E : 6.322 →
21.164, step 1.322 GeV. This is an outcome of Exercise 11.2.

11.3 Exercises

The first exercise deals with EMMA lattice and ring. It is concluded with a 6D bunch
acceleration simulation.
The second exercise deals with the DF cell of the second, high energy, eRHIC
21 GeV ERL recirculating loop (FFAG2). Simulations include synchrotron radiation
and spin diffusion.
11.1 EMMA Ring
Solution 11.1
In this exercise EMMA cell and 42-cell ring (Fig. 11.3) input data files are built
(after [26, 27]), their optical parameters are produced. Accelerating gaps are installed
and a rapid acceleration cycle is simulated.
Figure 11.13 displays a synoptic of EMMA ring, an outcome of the present
exercise.
(a) Construct EMMA cell. Parameters are given in Table 11.1.
Consider three different simulations of the quadrupoles:

– use MULTIPOL and hard-edge model;

– use MULTIPOL accounting for fringe fields, take MULTIPOLE[XE.= XS.= 5 cm,
.λ E = λ S = 2 cm] (. X E,S ≫ λ E,S to ensure that the magnetic field at the extrem-

Fig. 11.13 EMMA ring and

its injection and extraction
kicker-septum systems, and
the 15 MeV orbit (thick line
in the quadrupoles, blue). A
graph obtained using zpop,
menu 7: 1/1 to open
zgoubi.plt; 2/[42, 48] for
. X, Y laboratory coordinates;
7 to plot
458 11 FFAG, Linear

Table 11.1 Parameters of EMMA ring, a 42 edge polygon. The cell is straight, the .360/42 deg
polygon corners are at the interface between long drift and QD quadrupole
Energy range MeV 10–20
Number of turns .<16

Circumference m 16.568
Lattice D/F doublet
No of cells 42
RF frequency and range GHz; MHz 1.3; 5.6
No of RF cavities 19
RF voltage kV/cavity 20–120
EMMA cell:
– length cm 39.448
– drifts, short/long cm 5/21
– length, QD/QF cm 7.57/5.878
– gradient, QD/QF T/m 4.704/6.695
– QD/QF offsets cm 3.4048/0.7514

ity of the fall-offs (where it is truncated) is negligible compared to the field in

quadrupole body);
– use DIPOLES[. R M ≈ 105 ], with both quadrupoles comprised within
DIPOLES[AT] aperture so to allow linear superimposition (addition) of fringe
fields in the overlapping region.
For these three different models, repeat the following: plot the evolution with
energy . E : 10 → 20 MeV, step 1 MeV, of
– closed orbits in laboratory frame; field along the latter;
– betatron functions and dispersion along the cell;
– tunes;
– time-of-flight.
Use FIT embedded in REBELOTE[IOPT.= 1] do-loop for the closed orbit energy
scan. Add MATRIX[IORD.= 1, IFOC.= 11] within the loop (just before REBE-
LOTE) to get tunes and periodic beam matrix at each loop.

(b) Give phase space diagrams of the maximum stable amplitudes, horizontal and
vertical.
(c) Serpentine acceleration.
Construct EMMA ring—at this stage Fig. 11.13 can be produced. Add one accel-
erating cavity per cell, using CAVITE[IOPT.= 7] (note that EMMA only had 19
cavities, as two series of three cells were taken for the injection system: two kick-
ers and a septum, and the extraction system: a septum and two kickers [26, 27]).
From Sect. 11.2.2 figure a proper RF voltage for acceleration from 10 to 20 MeV in
.∼ 15−20 turns.
11.3 Exercises 459

Table 11.2 Parameters of a DF quadrupole doublet cell of eRHIC ERL FFAG2 return loop [12,
Appendix A]. Both BD and QF are offset quadrupoles, treated here as combined function dipoles
Top recirculation energy GeV 21.164
Energy of first recirculation GeV 6.622
Cell length m 3.3624
Number of cells per sextant 120
Cell structure, in that order:
Upstream drift length m 0.09652479
BD combined function dipole:
– length m 1.129301
– dipole field T 0.00293364
– gradient T/m .−0.5225857

Middle drift length m 0.19304957

QF quadrupole:
– length m 1.847002
– dipole field T 0.00293364
– gradient T/m 0.3728876
Downstream drift length m 0.09652479

Simulate an acceleration cycle of a hollow bunch (i.e., all electrons on a common

invariant; OBJET[KOBJ.= 8] can be used for that) to 20 MeV, followed by deceler-
ation back to 10 MeV. Produce the turn-by-turn phase space trajectory. Storage of
turn-by-turn data can use FAISTORE.

11.2 eRHIC ERL FFAG2 loop. Synchrotron Radiation. Polarization

Solution 11.2
The eRHIC project of a linac-ring electron-ion collider (Fig. 11.4) is based on the
existing RHIC, and on an electron ERL comprised of two linear FFAG recirculation
loops (FFAG1 and FFAG2) alongside RHIC ring. The ERL takes an electron bunch
through a 1.322 GeV linac, up to 21.164 GeV and back down to 1.3 GeV for energy
recovery [11, 12].
In this simulation exercise the 6.322–21.164 GeV FFAG2 recirculating loop is
considered. Optical properties of the quadrupole doublet cell are produced, as well
as effects of synchrotron radiation on electron dynamics and on polarization. More
simulations and their outcomes can be found in [11, 12].
In a similar periodicity to RHIC, the electron ERL return loop is comprised of 6
sextants, interleaved with 6 long straights. Each sextant is comprised of 120 cells.
One of the straights includes a 1.322 GeV linac.
(a) Build the DF quadrupole doublet cell input data file, following its parame-
ters given in Table 11.2. Use MULTIPOL and a hard-edge model to simulate both
quadrupoles.
(b) Consider the cell. Produce graphs of the evolution with recirculation energy
. E : 6.322 → 21.164 GeV, step 1.322 GeV, of:
460 11 FFAG, Linear

– periodic orbits in BD and QF quadrupoles;

– field along the latter;
– time-of-flight;
– tunes and chromaticities;
– bend-plane components of spins across BD and QF (assumed entering the cell
with .S = Sl ).

Use FIT embedded in REBELOTE[K.= 0, IOPT.= 1] do-loop for that closed

orbit energy scan. Add MATRIX within the loop to get tunes and periodic beam
matrix. MATRIX[PRINT] can be used to log MATRIX computation outcomes in
zgoubiMATRIX.out.
Produce graphs of the betatron functions and dispersion along the cell, at 6.622
and 21.164 GeV.
(c) Consider FFAG2 recirculation loop, assumed here only comprised of six arcs
(no long straights), 120 cells each. Introduce synchrotron radiation, the switch for
that is SRLOSS[KWRD.= MULTIPOL].
Note: field scaling as discussed in Sect. 9.2.5, which adapts beam line strength to
upstream SR loss, is not considered here. The reason is that, (i) the present FFAG
beam line encompasses 12 different energies, and (ii) the effect, orbit spiraling mostly
affecting the highest energy, 21.164 GeV, is within transverse acceptance.
Produce a graph of the evolution of energy loss per particle, and of the rms energy
of the radiated photons, as a function of energy, from 6.322 to 21.164 GeV, over the
recirculation loop.
Check against theoretical expectations.
(d) Add spin, all 5,000 longitudinal at start. Produce a graph of the evolution of the
average and rms values of vertical spin angle with energy, over the 12 recirculations
from 6.322 to 21.164 GeV.
Check against theoretical expectations.
Give the evolution with energy of the horizontal and longitudinal emittances.

11.3 eRHIC 22 GeV ERL, From Start to End

Solution 11.3
This simulation recirculates a beam 23 times in the single linear FFAG2 loop,
via spreaders, combiners and linac. Beam energy increases first, from 6.622 to
21.164 GeV, and decreases next, back down to 6.622 GeV. It is a good idea here,
to refer to [12, 31].
(a) Simulation of eRHIC from start to end requires building its 120 m, 1.322 GeV
linac. The latter is comprised of 42 modules. A module is approximated as DRIFT
[L.= 53.249249 cm]-CAVITE[K.= 0, IOPT.= 10]-DRIFT[L.= 53.249249 cm]. CAV-
ITE[K.= 0, IOPT.= 10] is a point transform, yet its data include length so to allow
updating the distance and time of flight of particles [28, cf. CAVITE]. Parameters are
as follows: length 1.7749 m; RF 422.26 MHz; peak voltage 31476190 V; RF phase
.π/2; matrix model option +1 [[28] cf..CAVITE[IOPT = 10, IOP = +1]]. With a

peak voltage of 31.47619 MVolts and synchronous phase of .π/2, 42 modules total
1.322 GeV.
11.4 Solutions of Exercises of This Chapter: FFAG, Linear 461

Distribute a few tens of electrons on same value horizontal and vertical invariants
ϵ = ϵ Z = 25 π μm, normalized. Orient the invariant for a waist in the middle of
. Y
the linac, both planes. MCOBJET[KOBJ.= 3] can be used. Transport these particles
through the linac, produce graphs of the initial and final invariants, horizontal and
vertical.
(b) This start to end eRHIC ERL simulation uses the results of the previous
exercise. When set up, it takes the beam from 6.622 upto 21.164 GeV, and then back
down to 6.622 GeV.
The simulation requires the use of the keywords

– REBELOTE[IOPT.= 1] in combination with GOTO[PASS#] and GOTO

[GOBACK] to switch the beam to proper subsystems (return loops, spreader,
combiner and linac),
– REBELOTE[IOPT.= 1, LMNT.= DRIFT[DltaPhase]] to flip the RF phase by pi
at entrance to the linac, for energy recovery from pass number 14 on,
– INCLUDE in abundance, so to allow separate data files for the recirculating chan-
nel(s), spreaders, combiners and linac—which has the additional merit of simpli-
fying the main input data file,
– FINISH, which ends the job, in lieu of END which cannot be used in this context.
– AUTOREF possibly, for fine-centering of the beam at entrance of the sub-systems.

Moreover, spreaders and recombiners input data files have to be set up. They can
be found as modules in https://sourceforge.net/p/zgoubi/code/HEAD/tree/branches/
exemples/didacticExercises/LR-eRHIC/upDown.dat file.

11.4 Solutions of Exercises of This Chapter: FFAG, Linear

11.1 EMMA Ring

(a) EMMA Cell and optical parameters.
.• For a hard-edge model of the quadrupoles: use MULTIPOL[. X E = X S = 0,
.λ E = λ S = 0]. The input data file is given in Table 11.3.

.• For a model of the quadrupoles accounting for fringe fields: use MULTI-

POL[. X E = X S = 5 cm, .λ E = λ S = 2 cm] (the fringe extents .λE,S are taken com-
mensurate with quadrupole bore diameter). Thus, in Table 11.3 replace

0. 0. 1.00 1.00 1.00 1.00 1.00 1. 1. 1. 1.

by

5. 2. 1.00 1.00 1.00 1.00 1.00 1. 1. 1. 1.

.• A model of the cell using DIPOLES[. R M → ∞]: The input data file can be
found in [26, Table 3], and the magnetic field it yields in [26, Fig. 5].
462 11 FFAG, Linear

Table 11.3 Input data file: EMMA cell EMMACell.inc. This file defines the double-cell segment
#S_cell to #E_cell for use in further questions. Run as it is, it computes the first order mapping at
11 different energies, from 10 to 20 MeV, step 1 MeV

With either one of these three different models: use FIT embedded in REBELOTE
do-loop [28, cf. REBELOTE], to find the periodic .Y , .T horizontal conditions at cell
end, for 11 different energies from 10 MeV to 20 MeV. MATRIX[PRINT] stacks
matrix computation outcomes in zgoubi.MATRIX.out.
Raytracing outcomes are displayed in Fig. 11.14, 11.15, 11.16 and 11.17. Opti-
cal functions, Fig. 11.16, are computed using the [pathTo]/zgoubi-code/toolbox/beta
FromPlt tool found in zgoubi package [29].
Completing these results accounting for all three models is left to the reader.
Similar cases may be found in various examples in zgoubi sourceforge repository,
https://sourceforge.net/p/zgoubi/code/HEAD/tree/branches/exemples/ folder.
11.4 Solutions of Exercises of This Chapter: FFAG, Linear 463

Fig. 11.14 Closed orbits along a pair of cells. A graph obtained using zpop: menu 7; 1/1 to open
zgoubi.plt; 2/[42,48] for . X lab versus .Ylab ; 7 to plot

Fig. 11.15 Field along

closed orbits, in the fringe
field model. A graph
obtained using zpop: menu
7; 1/1 to open zgoubi.plt;
2/[6,32] for . B Z versus .s;
7 to plot

The periodic beam matrix at injection is needed in question (c). Its value out of
the first step in the REBELOTE scan is:

(b) Maximum stable amplitudes.

They are computed using the [pathTo]/zgoubi-code/toolbox/searchStabLim tool
found in zgoubi package [30]. Results are given in Figs. 11.18 and 11.19.
(c) A serpentine acceleration-deceleration cycle.
An input data file for this simulation is given in Table 11.4: a 300 electron bunch
is tracked over 17 turns at an acceleration rate of 68 kV per cavity. Particle data are
logged turn by turn in zgoubi.fai. Tracking outcomes are displayed in Fig. 11.20.
464 11 FFAG, Linear

Fig. 11.16 Betatron functions and dispersion along a cell (with some artifacts in overlapping fringe
field regions)

Fig. 11.17 Tunes and time-of-flight parabola as a function of particle energy

11.2 eRHIC ERL FFAG2 loop. Synchrotron Radiation. Polarization

(a) Input data file for eRHIC FFAG2 DF cell [12].
The input data file, constructed from the parameters of Table 11.2, is given in
Table 11.5. The orbital angle through a cell is explicit, as MULTIPOL[ALE] param-
eter, thus the total deviation for 6 arcs, i.e., times 120 cells, is

.720 × 2 × ( 1.6564835 × 10−3 + 2.7090305 × 10−3 ) ≈ 2π rad

-ALE in BD -ALE in QF
11.4 Solutions of Exercises of This Chapter: FFAG, Linear 465

Fig. 11.18 Maximum stable

horizontal amplitudes, from
inner to outer invariant at 10,
13, 15, 17 and 20 MeV

Fig. 11.19 Maximum stable

vertical amplitudes, from
inner to outer invariant at 10,
13, 15, 17 and 20 MeV

in agreement with the present working hypotheses (things are a little different in the
actual ring design, which is comprised of six 102 cell arcs, dispersion suppressor
FFAG sections between arcs [11], and merger and combiner sections at linac ends).
It is a good idea to first check in zgoubi.res the length of the cell, as obtained when
running this file.
(b) Optical parameters of the DF cell.
Scans take care of producing cell parameters for each one of the 12 passes. A typi-
cal input data file is given in Table 11.6, it loops on MATRIX computation, so produc-
ing a scan of the wave numbers and other parameters, logged in zgoubi.MATRIX.out.
The latter is read and plotted using a gnuplot script (Table 11.6). Results are given in
Fig. 11.21.
Use OBJET[KOBJ.= 6] and MATRIX[IORD.= 2, IFOC.= 11] to add computation
of chromaticities. Or as well, use a TWISS command, however it takes longer, as it
does more than MATRIX.

Periodic orbits: OPTIONS[.plt, 2] (Table 11.6) causes log of stepwise parti-

cle data across the quadrupoles, in zgoubi.plt. These data include periodic orbit
466 11 FFAG, Linear

Table 11.4 Left input data file, EMMA2Cells+CAV.inc: EMMA double cell and accelerating
cavity. This file defines the cell segment #S_2cells-withCAV to #E_2cells-withCAV INCLUDEd
in the complete ring, right file. Right input data file: a 300 electron bunch accelerated-decelerated
in EMMA ring

Fig. 11.20 Longitudinal

phase space of an
acceleration-deceleration
cycle over .10 ≤ E ≤
20 MeV range in EMMA
ring. Turn-by-turn bunch
records over 17 turns are
shown here. A graph
obtained using zpop: menu
7; 1/5 to open zgoubi.fai;
2/[18, 19] for . E versus RF
phase; 7 to plot

coordinates, as well as magnetic field along the latter. To get graphs, a pos-
sibility is to use gnuplot scripts from zgoubi toolbox in sourceforge, in the
folder [pathTo]/zgoubi-code/toolbox/gnuplot_Zplt. . B Z (s) for instance uses gnu-
plot_Zplt_sBZ.gnu. Trajectories use gnuplot_Zplt_sYZ.gnu. The gnuplot commands
are reproduced in Table 11.6. Expected outcomes are given in Fig. 11.22.
Betatron functions: a similar question is treated in the previous exercise (cf.
Fig. 11.16), apply the same method here. The expected result is displayed in
Fig. 11.23.
11.4 Solutions of Exercises of This Chapter: FFAG, Linear 467

Table 11.5 Input data file: eRHIC DF quadrupole doublet cell. As it is, it produces a transport
matrix and the periodic beam matrix for the DF cell, at 6.622 GeV. This file defines the segment
S_eRHIC_BDCell:E_eRHIC_BDCell for INCLUDEs in subsequent input data files

Fig. 11.21 A scan of cell tunes.νY (square markers) and.ν Z (circles) for the 12 recirculated energies
in .6.622 → 21.164 GeV eRHIC FFAG2 loop. The solid line is to guide the eye
468 11 FFAG, Linear

Table 11.6 Input data file: a do-loop scan on MATRIX computation. Outcomes are logged
in zgoubi.MATRIX.out (a consequence of MATRIX[PRINT]). This file INCLUDEs the
S_eRHIC_BDCell:E_eRHIC_BDCell segment grabbed from the input data file of Table 11.5
11.4 Solutions of Exercises of This Chapter: FFAG, Linear 469

Fig. 11.22 Left, red curves: transverse excursion of the 12 periodic orbits across the FFAG cell
quadrupoles (vertical is null). Right: magnetic field experienced along these 12 orbits, in a hard-
edged model

Fig. 11.23 Optical functions at 6.622 GeV (left) and 21.164 GeV (right), from stepwise ray-tracing
across FFAG2 cell

Spin components: add PARTICUL[POSITRON] and SPNTRK[KSO.= 1] after

OBJET, to define initial spin coordinates. They will be tracked and logged in
zgoubi.plt. . S X , SY , and . S Z are respectively columns 33, 34 and 35 [28, Sect. 8.3].
(c) SR energy loss.
The input data file for this simulation is given in Table 11.7. The way it works:
– MCOBJET[KOBJ.= 3] defines a 6D bunch, comprised of 5,000 electrons, with
null 6D emittance here. Due to REBELOTE[K.= 0], MCOBJET generates a new
set of initial coordinates (namely, the periodic orbit coordinates Y and T for a new
D value) at the start of each new pass
– defining PARTICUL is necessary as mass and charge are needed for computation
of SR, and for integration of spin motion
– the cell is INCLUDEd 720 times, that makes up the ring
– REBELOTE loops 12 times, that takes care of the 12 passes (the lowest energy
happens to be treated twice, not an issue—REBELOTE could as well start from
the second energy for 11 passes, instead)
– prior to launching the next pass, REBELOTE[NPRM.= 45, 40, 41] changes the
relative reference momentum of the particles (D parameter, number 45 under
470 11 FFAG, Linear

Table 11.7 Input data file: transport a 5000 electron bunch, accounting for SR energy loss. Initial
6D emittance is null. The ring is comprised of 720 cells (no long straights)

MCOBJET) and the periodic coordinates (Y and T, number 40 and 41 under

MCOBJET) accordingly. The latter are taken from the outcomes of question (b).

The informations of interest happen to be in zgoubi.res—no need of any other

log file. Namely, “Average energy loss per particle per pass”,
and “Average energy of radiated photon”, found under SRPRNT. A
graph is displayed in Fig. 11.24, obtained using a gnuplot script given in Table 11.7.
Checking against theoretical expectations is left to the reader.
11.4 Solutions of Exercises of This Chapter: FFAG, Linear 471

Fig. 11.24 Energy dependence of the average energy loss (left axis, solid squares) and rms energy of
radiated photon (right axis, empty circles), over the FFAG2 recirculation loop, at the 12 recirculated
energies. The solid lines are to guide the eye

FAISCEAU computes the concentration ellipses, for 5,000 particles here. Thus
graphs of the emittance growth due to synchrotron radiation can be obtained in
a similar way to the SR loss graph (cf. gnuplot script, Table 11.7), by a grep of
‘(Y, T)’, ‘(Z, P)’ and ‘(t, K)’ as found under FAISCEAU in zgoubi.res, for instance
(case of last, 21.164 GeV recirculation):

(d) Adding spin.

SPNTRK[KSO.= 1] (initial spin along the X axis, longitudinal) is included in the
input data file used in (c), Table 11.7. Running it also tracked the spins. Particle and
spin data are logged in zgoubi.fai, as a result of FAISTORE command.
Thus, it just remains to produce a graph from the previous question outcomes.
There are various ways to compute the rms spin angle from the 5000 electron sample
at the end of the loop. An ancillary program can be used, awk commands in a
gnuplot script is another possibility. The outcome is displayed in Fig. 11.25, which
also displays histograms of the longitudinal spin component at 3 different passes, 6,
8 and 11, accounting for an initial vertical jitter.

11.3 eRHIC 22 GeV ERL, From Start to End

It is a good idea here, to refer to [11, 12].
(a) eRHIC linac.
eRHIC 120 m, 1.322 GeV linac is essentially based on CAVITE. CAVITE
[IOPT.= 10] transports particles using a Chambers matrix [28, cf. CAVITE]. Trans-
verse damping can be accounted for, or not, using the . I O P = ±1, ±2 flag.
472 11 FFAG, Linear

Fig. 11.25 Left: polarization loss due to energy spread, from tracking (markers) and from Eq. 11.9
(dashed line). Right: an histogram of the vertical spin angle at pass 6, 8 and 11, in the presence of
an initial .δ Z = 0.1 mm vertical beam jitter

Table 11.8 Input data files. Left: linacModule.inc, defines a module of eRHIC linac. Right:
linac.INC.dat, defines the 42 module, 120 m long 1.322 GeV linac

eRHIC linac simulation input data file is given in Table 11.8. Tracking results are
displayed in Fig. 11.26.
(b) eRHIC, from start to end.
Guidance in setting up the input data file can be found in an existing com-
plete recirculating linac simulation, an energy recovery linac (ERL), 12 passes
up 11 passes down, in https://sourceforge.net/p/zgoubi/code/HEAD/tree/branches/
exemples/didacticExercises/LR-eRHIC/ folder. This ERL simulation is based on the
keywords:
References 473

Fig. 11.26 An initial 1000-particle bunch with all transverse particle coordinates taken on a given
invariant .γ u 2 + 2αuu ' + βu '2 = ϵu /π is tracked from entrance to exit of the linac (u stands for Y
or Z;.ϵY = ϵ Z = 25 π μm, normalized, 6.622 GeV here). The figure shows horizontal (left plot) and
vertical (right plot) phase spaces at linac entrance (converging ellipse) and exit (diverging ellipse)

– REBELOTE[IOPT.= 1] in combination with GOTO[PASS#] and GOTO

[GOBACK] to switch the beam to proper subsystems (return loops, spreader,
combiner and linac),
– REBELOTE[IOPT.= 1, LMNT.= DRIFT[DltaPhase]] to flip the RF phase by pi
at entrance to the linac, for energy recovery from pass number 14 on,
– abundant use of INCLUDE so to allow separate data files for the recirculating
channel(s), spreaders, combiners and linac—which has the merit additional of
simplifying the main input data file,
– FINISH, which ends the job, in lieu of END which cannot be used in this context.

References

1. F.E. Mills, C. Johnstone, Proceedings of the 4th International Conference on Physics Potential
& Development of .μ+ − μ− Colliders, San Francisco, CA (UCLA, Los Angeles, CA, 1999),
pp. 693–698
2. C. Johnstone, et al., Fixed field circular accelerator designs, in Proceedings of the 1999 Par-
ticle Accelerator Conference, New York (1999). https://accelconf.web.cern.ch/p99/PAPERS/
THP50.PDF. Fig. 11.1: Copyrights under license CC-BY-3.0, https://creativecommons.org/
licenses/by/3.0; no change to the material
3. S. Koscielniak, C. Johnstone, Longitudinal dynamics in an FFAG accelerator under conditions
of rapid acceleration and fixed, high RF, in Proceedings of the 2003 Particle Accelerator
Conference. https://accelconf.web.cern.ch/p03/PAPERS/TPPG009.PDF
4. D. Trbojevic, et al., Fixed field alternating gradient lattice design without opposite bend,
in Proceedings of EPAC 2002, Paris, France. https://accelconf.web.cern.ch/e02/PAPERS/
WEPLE051.pdf
5. J.S. Berg, et al., Cost-effective design for a neutrino factory. Phys. Rev. Spec. Topics - Accel.
Beams 9, 011001 (2006). https://journals.aps.org/prab/pdf/10.1103/PhysRevSTAB.9.011001
6. S. Machida, et al., What we learned from EMMA, in MO2PB01 Proceedings of Cyclotrons
2013, Vancouver, BC, Canada
7. R. Edgecock, EMMA - the world’s first non-scaling FFAG, in THOBAB01 Proceedings of
PAC07, Albuquerque, New Mexico, USA
474 11 FFAG, Linear

8. S. Machida, et al., Acceleration in the linear non-scaling fixed-field alternating-gradient accel-

erator EMMA. Nat. Phys. 8, 243–247 (2012). https://www.nature.com/articles/nphys2179
9. J.S. Berg, The EMMA lattice NIM A 596, 276–284 (2008)
10. D. Trbojevic, et al., ERL with non-scaling fixed field alternating gradient lattice for eRHIC,
in IPAC2015, Richmond, VA, USA. https://accelconf.web.cern.ch/IPAC2015/papers/tupty047.
pdf
11. F. Méot, et al., Tracking studies in eRHIC energy-recovery recirculator. Tech Note C-
A/eRHIC/45, BNL C-AD, Upton, LI, NY 11973 (2015). https://www.osti.gov/biblio/1210189
12. F. Méot, et al., eRHIC ERL modeling in zgoubi. Tech Note C-A/eRHIC/49, BNL C-AD, Upton,
LI, NY 11973 (2016). https://technotes.bnl.gov/PDF?publicationId=38865. https://www.osti.
gov/biblio/1335396
13. D. Trbojevic, CBETA Technical Report. Tech Note BNL-211697-2019-TECHCBETA/031
(2018). https://www.osti.gov/servlets/purl/1515414
14. C. Mayes, CBETA Multipass lattice design, in Proceedings of the ERL Workshop, CERN, June
20 (2017). https://accelconf.web.cern.ch/erl2017/talks/tuidcc003_talk.pdf. Fig. 11.5: Copy-
rights under license CC-BY-3.0, https://creativecommons.org/licenses/by/3.0; no change to
the material
15. F. Méot, et al., Simulation of the CBETA 4-pass FFAG ERL using field-maps - exclusively. Tech
Note BNL-211811-2019-JAAM, BNL (2019). https://www.osti.gov/servlets/purl/1529889
16. S.A. Bogacz, et al., 20-24 GeV FFA CEBAF energy upgrade, in 12th International Parti-
cle Accelerator Conference IPAC2021, Campinas, SP, Brazil. https://accelconf.web.cern.ch/
ipac2021/papers/mopab216.pdf
17. F. Méot, N. Tsoupas, S. Brooks, D. Trbojevic, Beam dynamics validation of the halbach tech-
nology FFAG cell for cornell-BNL energy recovery Linac. Nuclear Inst. Methods Phys. Res.
A 896, 60–67 (2018)
18. F. Lemuet, Collection and muon acceleration in the neutrino factory project. Ph.D. Thesis
dissertation, CEA and CERN, Paris-Saclay University (2007). https://inis.iaea.org/collection/
NCLCollectionStore/_Public/42/013/42013892.pdf
19. https://cds.cern.ch/collection/Neutrino%20Factory%20Notes?as=1&ln=ru
20. https://www.bnl.gov/ffaworkshop/events/
21. E. Keil, A.M. Sessler, Muon acceleration in FFAG rings. CERN-NUFACT-note-139, CERN-
AB-2004-033 (ABP) (2004). https://cds.cern.ch/record/784564/files/nufact-note-139.pdf
22. J.S. Berg, Amplitude dependence of time of flight and its connection to chromaticity. NIMA
570(1), 15–21 (2007)
23. G. Leleux, et al., Synchrotron radiation perturbations in long transport lines, in Proceeding
of the PAC 1991 Accelerator Conference. https://accelconf.web.cern.ch/p91/PDF/PAC1991_
0517.PDF
24. F. Méot, V. Litvinenko, T. Roser, Polarization transport in the ERL-ERL FCC e+ e- Collider.
Tech Note C-A/AP/684 BNL C-AD (2022)
25. V. Ptitsyn, Electron polarization dynamics in eRHIC, in EIC’14 Workshop. http://appora.fnal.
gov/pls/eic14/agenda.full
26. F. Méot, Y. Giboudot, et al.: Beam dynamics simulations regarding the experimental FFAG
EMMA, using the on-line code. THPD023, Proceedings of IPAC’10, Kyoto, Japan. https://
accelconf.web.cern.ch/IPAC10/papers/thpd023.pdf
27. F. Méot, The ray-tracing code Zgoubi - status. Nuclear Instrum. Methods Phys. Res. A 767,
112–125 (2014)
28. F. Méot, Zgoubi users’ guide. https://www.osti.gov/biblio/1062013-zgoubi-users-guide An
up-to-date version of the guide: https://sourceforge.net/p/zgoubi/code/HEAD/tree/trunk/
guide/Zgoubi.pdf
29. From zgoubi toolbox, part of the sourceforge package: a Fortran tool to compute
optical functions from a zgoubi.plt output file, and some related gnuplot scripts, see
the README and example, there: https://sourceforge.net/p/zgoubi/code/HEAD/tree/trunk/
toolbox/betaFromPlt/
References 475

30. From zgoubi toolbox, part of the sourceforge package: a Fortran tool to perform a dynamic
aperture scan, and some related gnuplot scripts, see the README and example, there: https://
sourceforge.net/p/zgoubi/code/HEAD/tree/trunk/toolbox/searchStabLim/
31. F. Méot, et al., Tracking studies in eRHIC energy-recovery recirculator. Tech Note C-
A/eRHIC/45, BNL C-AD, Upton, LI, NY 11973 (2015). https://www.osti.gov/biblio/1210189

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0
International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing,
adaptation, distribution and reproduction in any medium or format, as long as you give appropriate
credit to the original author(s) and the source, provide a link to the Creative Commons license and
indicate if changes were made.
The images or other third party material in this chapter are included in the chapter’s Creative
Commons license, unless indicated otherwise in a credit line to the material. If material is not
included in the chapter’s Creative Commons license and your intended use is not permitted by
statutory regulation or exceeds the permitted use, you will need to obtain permission directly from
the copyright holder.
Chapter 12
Beam Lines

Abstract This chapter introduces beam transport and manipulations in beam lines.
It provides the theoretical material resorted to in the simulation exercises, leaning on
charged particle optics and beam manipulation concepts introduced in earlier Chap-
ters. The simulation of beam lines and specific functionalities they ensure require new
optical elements, such as WIENFILTER, EBMULT, high order multipoles, etc. Par-
ticle monitoring resorts to keywords introduced in the previous Chapters, including
FAISCEAU, FAISTORE, possibly PICKUPS, and some others. Spin motion com-
putation and monitoring resort to SPNTRK, SPNPRT, FAISTORE. Optics matching
and optimization use FIT[2]. INCLUDE is sometimes resorted to, mostly in order
to modularize and/or simplify the input data files. SYSTEM is used to, mostly, call
gnuplot scripts so as to end simulations with some specific graphs, or animations.
These scripts read their data from output files filled up during the execution of the
code, such as zgoubi.fai (resulting from the use of FAISTORE), zgoubi.plt (resulting
from IL .= 2 flag), or other zgoubi.*.out files resulting from a PRINT command.

12.1 Introduction

Beam lines ensure diverse functions such as

– beam transfer between accelerators,
– injection into, and extraction from accelerators,
– beam transport to experimental areas, purification systems, mass separators,
– beam delivery to collision points,
– beam expansion and uniformization,
– microprobes, microscope columns,
– medical rotating gantries,

and more.
Moreover, in designing large accelerators, specific sections with specific function-
alities, “optical modules”, may be considered separately, and in doing so handled
as beam lines. This includes arcs, dispersion suppressors, spin rotators in polarized
electron storage rings, recirculating loops in energy recovery linacs, etc.

© The Author(s) 2024 477

F. Méot, Understanding the Physics of Particle Accelerators, Particle Acceleration
and Detection, https://doi.org/10.1007/978-3-031-59979-8_12
478 12 Beam Lines

Particle transport in beam lines in the Gauss approximation can be described using
elementary laws: parabolic, sinusoidal or hyperbolic. The complexity of a beam line,
in a first order approach, results from the variety of possible combinations of these
laws in the optical sequence: a particle will for instance describe an arc of a circle,
followed by a parabola and finally a sine or hyperbolic arc. This makes it possible to
design complex optical systems with a reduced number of basic optical elements.
The simulation exercises proposed in Sect. 12.3 use stepwise raytracing as this
provides detailed insight in fields and particle motion. Stepwise integration allows
dealing with optical elements featuring complex geometry and/or .E and/or .B field
structure, it allows for large transverse acceptance, large momentum offset, strong
field non-linearities, in an accurate manner.

12.2 Basic Concepts and Formulæ

In the following sections, a few different types of beam lines, with specific functional-
ities, are addressed. The underlying theory is introduced. Corresponding simulation
exercises complete the landscape.

12.2.1 Beam Expander

Beam uniformization is used for biological irradiation [1], at high power neutron
targets [2], it has been foreseen for hadrontherapy [3], radioactive waste treatment
and other material irradiation [4, 5]. Depending on the application the transverse
beam size needed may be in the centimeter (e.g., radio-biology) to meter range
(e.g., nuclear waste irradiation). Transverse uniformization is achieved by means of
octupoles (. B y | y=0 ∝ x 3 ), and possibly enhanced by adding higher odd-order compo-
nents (. B y | y=0 ∝ x 5 : dodecapole, …) [6–8]. The technique is generally implemented
in so-called beam expanders.
A two-dimensional beam expander, designed for nuclear waste irradiation, is
schemed in Fig. 12.1 [5]. It uniformizes the transverse density of a 700 MeV proton
beam with .0.6 µm emittances, over a typically .1 × 1 m2 area. Quadrupoles Q1–
Q7 ensure beam focusing at the irradiation target located further down the line at
.s = 37 m. Beam uniformization over a rectangular beam cross-section at the target
requires [7, 8]
(i) an non-linear lens OH (respectively, OV), for uniformization of the transverse
horizontal (respectively, vertical) particle density distribution;
(ii) flat horizontal and vertical phase-space beam ellipses at the location of, respec-
tively, the horizontal and the vertical lens;
(iii) the derivative .dy/dyl' of the polynomial in Eq. 12.2 (see Fig. 12.2 and notations
below) to change sign at least once.
12.2 Basic Concepts and Formulæ 479

Fig. 12.1 A beam expander,

including two non-linear
lenses OH and OV for
transverse beam
uniformization. Quadrupoles
Q4–Q7 control the size of
that rectangle. Quadrupoles
Q1–Q3 control the envelopes
at the non-linear lenses. The
graph displays the
√ linear
beam√envelopes . βx εx /π
and . β y ε y /π

Fig. 12.2 A geometrical

representation of transverse
density mapping, from a
Gaussian density upstream
of the lens (bottom), to
uniformized density at some
distance .s down the line (top
right; Eq. 12.3), via Eq. 12.2
mapping (top left;
'
. y(s) = λ1 (s) yl +

λ3 (s) yl + λ5 (s) yl' 5

' 3

here) [8]

Given hypothesis (i) above, uniformization can be tuned independently in each

plane. Hypothesis (ii) above requires .βl γl at the lens to be large (with .βl , αl =
−βl' /2, γl = (1 + αl2 )/βl the local optical functions at the lens). This results in the
beam ellipse at the non-linear lens to degenerate into a line, which can be written

βl γl ≫1
γ (s)y 2 + 2αl (s)yy ' + βl (s)y '2 = ε y /π
. l → yl = rl yl' (rl = −βl /αl ) (12.1)

In such conditions, the transverse position y(s) of a particle at arbitrary .s > sl down-
stream of the lens is related to its incidence angle . yl' at entrance of the lens by the
polynomial relationship
⎧
∑
n ⎨ n = 1 : octupole alone
λ2 p+1 (s) yl'
2 p+1
. y(s) = with n = 2 : octupole + dodecapole (12.2)
⎩
p=0 etc.
√ 2 p+1 √
wherein .λ1 (s) = rl β(s)/βl cos(∆φ), .λ2 p+1 (s) = −K 2 p+1 Lrl β(s)βl
sin(∆φ), .∆φ = φ(s) − φl is the phase advance from the lens, and . K 2 p+1 L ( p =
1 to n ≥ 1) are the integrated strengths of the .n odd-order non-linear components
present in the lens.
480 12 Beam Lines

Uniformized density
Let . f (yl' ) be the probability density of the angle variable . yl' at the lens. Thus the
density of the position variable . y at arbitrary .s > sl can be written
( ( ))
∑
N
f yl' i y(s)
g(y(s)) = |∑ )|| (12.3)
| n 2p(
.
i=1 | p=0 (2 p + 1) λ2 p+1 (s) yl' i y(s) |

wherein . yl' i .(i = 1, N ) are the . N ≤ 2n + 1 real roots of the polynomial Eq. 12.2
(to be found analytically for an octupole lens, n .= 1, numerically for octupole .+
dodecapole and beyond, .n ≥ 2).
The third hypothesis (iii) above ensures that the image .g(y) of . y by the Eq. 12.2
mapping presents a discontinuity of the first kind (a sufficient condition is.λ1 λ3 < 0).
In the case of an octupole lens it occurs at
( )
2 λ1 1/2
. ± y M = ± λ1 − (12.4)
3 3λ3

and in immediate neighboring if higher orders are added for improved uniformization.
Assuming that the incoming beam has a Gaussian density
( )
' 1 yl' 2
. f (yl ) = √ exp − 2 (12.5)
σ yl' 2π 2σ y '
l

an additional condition for near-uniform .g(y) is for the beam divergence .σ yl' =
√
γ y,l ε y /π to satisfy
'
.σ y ' ≲ yl (12.6)
l M

This ensures that the folding of the Gaussian projected beam density occurs in the
region . y ' ≈ σ yl' (Fig. 12.2).
Proper multipole lens strengths for that are

4 β(s) cos3 (φ(s) − φl )

.octupole: K3 L = (12.7)
2
27 y M βl2 sin(φ(s) − φl )

(K 3 L)2
added dodecapole:
. K5 L = − βl tan(φ(s) − φl ) (12.8)
4
Figure 12.3 displays the typical 2D-uniform beam cross-section so obtained at the
downstream end of the line, .s = 37 m (Fig. 12.1).
Non-linear beam envelopes
First order beam optics deals with envelopes in terms of the transport of√the optical
functions, and the .r ms beam size in dispersion free regions is .σ (s) = β(s)ε/π .
12.2 Basic Concepts and Formulæ 481

Fig. 12.3 Two-dimensional

uniform beam and projected
densities

Transverse apertures (chamber, collimators, etc.) are usually defined in terms of the
kσ envelope, with .k a few units depending on the loss tolerances. The parameter of
.
concern in this respect is the loss rate, or relative population beyond the .kσ trans-
verse excursion boundary, .τ = 1 − erf(kσ ). In a beam expander the non-linearities
introduced by the uniformization lens are so strong that it is no longer relevant to
address .τ in terms of the .kσ -envelope as the transverse density .g(y(s)) is way too
far from Gaussian. On the other hand, non-linear lenses may induce a substantial
change of the beam loss boundary compared to the linear envelope (Fig. 12.4).
Given some loss rate .τ (e.g., of the order of .10−7 − 10−9 in high power installa-
tions) [5], the .(2n + 1)-th order non-linear envelope .Y2n+1 (s) in a beam expander is
obtained by solving for .Y2n+1 (s) the integral equation
ʃ N ʃ
∑ yl' i (y=Y2n+1 (s))
Y2n+1 (s) | |
. g(y) dy ≡ f (yl' i ) |d yl' i | = 1 − τ (12.9)
0 i=1 yl' i (y=0)

Assuming a Gaussian incoming beam density (Eq. 12.5), this expression takes the
simpler form
∑
N
{ [ ]}
.sign(λ1 ) (−1)i erf yl' i Y2n+1 (s) = 1 − τ (12.10)
i=1

which can be solved numerically. Typical non-linear envelopes so obtained are dis-
played in Fig. 12.4.
A possible numerical procedure to transport non-linear envelops in that manner
is the following:
– first, the roots . yl' i (y = Y2n+1 (s)) of Eq. 12.2 are calculated (analytically in the
case .2n + 1 = 3, numerically if .2n + 1 ≥ 5),
– next, given .τ , Eq. 12.10 is solved numerically for .Y2n+1 (s).
That procedure also provides the local transverse density .g(y(s)) at arbitrary .s
along the beam line as illustrated in Fig. 12.5.
482 12 Beam Lines

Fig. 12.4 This graph shows

beam envelopes defined as a
.τ = 10
−7 loss boundary

(Eq. 12.10), horizontal [.x(s),

upper curves] and vertical
[. y(s), lower curves]. Three
cases are plotted: linear
envelopes (marker “1”); case
of octupole at OH and at OV
(“3”); case of
octupole+dodecapole at OH
and at OV (“5”) [8]

Fig. 12.5 Transverse

horizontal beam density
profiles .g(x) (a sub-product
of the transport of non-linear
envelopes using Eq. 12.10) at
various optical elements
along the beam line of
Fig. 12.4, given a sole
octupole component (.n = 1)
at OH [8]

12.2.2 Nano-Probe Final Focus Achromat

A magneto-electrostatic quadrupole combines electric and magnetic quadrupole

components, superimposed (Fig. 12.6) [9].
It presents the property that, in the case of non-relativistic beams, it can be made
achromatic to the second order so long as the electric and magnetic strengths

φ B
. Ke = and Km = (12.11)
a2W a Bρ

satisfy the relationship

. K m = −2K e (12.12)
12.2 Basic Concepts and Formulæ 483

Fig. 12.6 A principle sketch of a combined magnetic-electrostatic quadrupole lens. The conducting
pole tips (filled circles) are at 45.◦ to the magnetic pole tips—and at the same pole tip radius, here

QF QD
OBJET

IMAGE

Fig. 12.7 Principle scheme of a final focus second-order achromat for a nano-probe ion beam. It
is comprised of a pair of .E, B quadrupoles (Fig. 12.6) tuned to ensure same focal distance in both
transverse planes

There, .φ is the potential at the electrodes, .a is the inner radius at the electrodes and
at the magnetic poles, .W = p 2 / 2m is the kinetic energy of the particle of mass .m
and momentum . p, assumed non-relativistic, B is the magnetic field at pole tip. This
property of second order achromatism makes the doublet of .E, B quadrupoles an
option for a final focus doublet in a nano-probe [10] (Fig. 12.7).

12.2.3 A Muon Collect Channel

A Neutrino Factory accelerator complex [11] is aimed at the production of high flux
neutrino beams based on muon decay

μ± → e± + νe (νe ) + νμ (νμ )
.
484 12 Beam Lines

in a racetrack storage ring. A muon collect channel is part of the Neutrino Factory
design [12, 13]. Muon bunches are the product of pion decay

π →μ+ν
.

Pions bunches are obtained from multi-MW, GeV range proton bunches hitting a
production target. A funneling optical system downstream of a set of targets steers
the parent pion beams onto a common axis, into a long (a few tens of meters) muon
collect channel, which can be a solenoid, or a FODO lattice (see section “Neutrino
Factory Muon Collect Channel”). Following the muon collect channel, muon bunches
are cooled and then accelerated to multi-GeV energy for storage in a racetrack decay
ring. One of the long straight sections in the latter directs the neutrino beam towards
a far distant detector [11].
In a first part in the following, the momentum and time distributions of the parent
pion bunch as it decays, and of the muon bunch as it builds up, are examined. In a
second part a Funnel .+ FODO lattice design of the Neutrino Factory muon collect
channel is described.

Momentum and Energy Spectra

In-flight decay is sketched in Fig. 12.8 which also defines the scattering angles .θ
and .φ. Indices .π and .μ in the following designate respectively the pion and muon
particles. A superscript ‘.∗ ’ denotes quantities taken in the center of mass of the
two-body decay. Useful quantities in the following include,

Regarding pions:
– Momentum; energy; normalized velocity . pπ ; . E π ; .βπ
= pπ /E π
– Life time at rest; corresponding path length .τπ , = βπ γπ cτπ = pπ /η
.sπ
– Decay law . N (s) = N0 e
−ηs/ pπ (.η = m π / cτπ )
Regarding muons:
– Momentum; energy; normalized velocity . pμ ; . E μ ; .βμ= pμ /E μ
– Center-of-mass energy ∗
. Eμ = (m 2π( + m 2μ )/2m π
∗ ∗ ∗
)
– Energy in laboratory . E μ = γπ E μ + βπ pμ cos θμ

∗ 1 sin θμ∗
– Laboratory decay angles .φμ = φμ , tan θμ =
β
γπ ∗ + cos θμ∗
π
μ β

Parent Pion Bunch

The density of survived pions with initial momentum . pπ , as a function of flight
distance .s, writes
. gs| pπ (s, pπ ) = (η/ pπ ) exp (−ηs/ pπ ) (12.13)

Given .g pπ ( pπ ) the initial momentum density (at .s = 0), one gets the 2D density at
arbitrary .s > 0
12.2 Basic Concepts and Formulæ 485

Fig. 12.8 In-flight decay of

particle 1 at . M(x1 , y1 , z 1 ).
Particle 2 is emitted at angles
.θ, φ

ʃ ∞ ʃ
g
. s, pπ (s, pπ ) = gs| pπ × g pπ (and gs, pπ (s, pπ ) ds dpπ = 1) (12.14)
s=0

For the sake of simplicity a uniform initial pion momentum density may be consid-
ered,
{
1/( pπ2 − pπ1 ) if pπ ∈ [ pπ1 , pπ2 ]
. g pπ ( pπ ) = 1∆pπ ( pπ ) = (12.15)
0 otherwise

The resulting form of .gs, pπ (s, pπ ) is shown in Fig. 12.9, given a pion bunch launched
at .s = 0 with zero size and . pπ ∈ [100, 500] MeV/c. Integrating Eq. 12.14 with
respect to .s yields the . pπ -density of the decayed parent pions at distance .s,
ʃ [ ( )]
| s
ηs
. g pπ ( pπ )|s = gs, pπ (s, pπ ) ds = 1∆pπ ( pπ ) 1 − exp − (12.16)
0 pπ

The . pπ -density of the non-decayed pion population ensues,

| |
. g̃ pπ ( pπ )|s = (g pπ ( pπ ) − g pπ ( pπ )|s ) = 1∆pπ ( pπ ) exp (−ηs/ pπ ) (12.17)

The.s-dependence of the pion bunch average momentum follows, namely (Fig. 12.10)
ηs

∑ pπ2 i − η s pπi − η2 s 2 e pπi Ei(− pηπs )

ʃ (−1)i i
p g̃ pπ ( p)|s dp i=1,2
2 eη s/ pπi
. p̄π (s) = ʃ = ηs
g̃ pπ ( p)|s dp pπi + η s e pπi Ei(− pηπs )
∑
(−1)i i
i=1,2
eη s/ pπi
(12.18)
486 12 Beam Lines

Fig. 12.9 Left: pion bunch momentum density along the 30 m decay channel (Eq. 12.14), assuming
uniform initial momentum density over [100, 500] MeV/c (Eq. 12.15). Right: .s-sections of the
former, at various distances along the collect channel

Fig. 12.10 Average

momentum of pion and
muon bunches over 60 m
flight distance (Eq. 12.18)

Muon bunch

Given a fixed muon momentum . pμ ∈ [γπ (βπ E μ∗ − pμ∗ ), γπ (βπ E μ∗ + pμ∗ )], the decay
muon momentum (Fig. 12.11 ) satisfies a . pπ -conditional density which writes
| | mπ pμ mπ
. g pμ | pπ ( pμ ) = gθ ∗ (θμ∗ ) |dθμ∗ /dpμ | = ∗
/ = βμ (12.19)
2 pπ pμ p 2 + m 2 2 pπ pμ∗
μ μ

wherein (Fig. 12.11-right) .gθ ∗ (θμ∗ ) = sin θμ∗ /2 .(θμ∗ ∈ [0, π ]) is the decay angle den-
sity. Outside the . pμ interval .g pμ | pπ ( pμ ) ≡ 0. The muon energy density at fixed . pπ
writes
mπ
g
. E μ | pπ (E μ ) = with E μ ∈ [γπ (E μ∗ − βπ pμ∗ ), γπ (E μ∗ + βπ pμ∗ )] (12.20)
2 pπ pμ∗

A non-zero pion bunch momentum extent is accounted for by multiplying the

. pπ -conditional density .g pμ | pπ ( pμ ) (Eq. 12.19) by the muon density at .s at given . pπ ,
12.2 Basic Concepts and Formulæ 487

Fig. 12.11 Left: density .g pμ | pπ ( pμ ) (Eq. 12.19) for initial pion momenta . pπ = 100, 300 and
500 MeV/c. Right: geometrical representation of its build-up from.gθ ∗ (θμ∗ ) in the change of variable
∗
.θμ → pμ

Fig. 12.12 Left: muon momentum density along the decay channel (from an analytical primitive
of the integral in Eq. 12.21 [13]). Right: .s-sections of the former, at various distances along the
collect channel, from rough numerical computation of the integral

. gs, pπ (s, pπ ) (Eq. 12.14). (The muon decay is not taken into account in the following
for simplicity, doing so would mean accounting for an .s-dependent muon survival
additional factor.) This yields the muon momentum spectrum at .s under the integral
form
ʃ ʃ s
. g pμ ( pμ )|s = g pμ | pπ dpπ gs, pπ (s, pπ ) ds ( lim g pμ ( pμ )|s = 1) (12.21)
∆pπ 0 s→∞

The .∆pπ integration interval is a function of . pμ following the dependence given in

Eq. 12.19 (not all pions can produce a muon of momentum . pμ ). A similar integral
holds for the energy spectrum .g Eμ (E μ )|s , given .g Eμ | pπ (Eq. 12.20).
The summation in Eq. 12.21 can be viewed as a superposition of the fixed-. pπ
muon spectra of Fig. 12.11, this is the way the muon spectra shown in Fig. 12.12
have been numerically calculated.
The determination of the mean momentum . p̄μ (s) and energy . Ē μ (s) of the muons,
and as well the momentum of the center of gravity of the hybrid bunch follow similar
488 12 Beam Lines

Fig. 12.13 Muon bunch time-kinetic energy density (Eq. 12.22, arbitrary units) at .s = 30 m, in the
case . pπ ∈ [200, 400] MeV/c

calculations as for the pion bunch, above, they are represented in Fig. 12.10. Average
momenta of both the pion and the muon bunches are increasing functions of the
distance, because the lower energy parent pions decay faster, whereas the average
momentum of the .π + μ bunch decreases monotonically (Fig. 12.10), a behavior
that can be accounted for to maintain constant focusing strength in tuning the decay
channel.
Muon bunch longitudinal phase space
The . pπ -conditional time density .gtμ | pπ ( pμ ) of the muons at arbitrary .s is derived
from .gθ ∗ (θμ∗ ) through a change of variable

θ∗ →
. μ tμ = sπ /cβπ + (s − sπ )/cβμ

This yields the time-energy two-dimensional density under the integral form
(Fig. 12.13)
ʃ s ʃ
g
. tμ ,E μ (tμ , E μ )|s = gtμ | pπ (tμ ) gs, pπ (sπ , pπ ) dpπ dsπ (12.22)
sπ =0 ∆pπ

The time boundaries of the distribution, namely,

(E 2∗ ± βπ pμ∗ )
ctμ± = (s − sπ )
. (12.23)
(βπ E 2∗ ± pμ∗ )
12.2 Basic Concepts and Formulæ 489

correspond to the fastest and slowest muon emitted by respectively the fastest and
slowest pion. The first two moments of .gtμ (tμ )|s can be calculated so as to derive
capture efficiencies as was done for the energy spectra.
The muon time density .gtμ (tμ )|s (Eq. 12.22) has an explicit dependence on .s. Note
that, given the pion energies in concern here, the flight distance .s can be considered
in good approximation as the position along the channel length.
The muon population at arbitrary .s can also be reconstructed from Eq. 12.22: the
.(tμ , E μ ) space can be meshed, . N0 gtμ ,E μ (tμ , E μ )|s ∆pμ ∆E μ gives the local number of
points on the mesh. Monte Carlo simulations of longitudinal phase-space at distance
.s along a drift confirm these results (see Exercise 12.10).

Neutrino Factory Muon Collect Channel

A target system of a neutrino factory requires the incident proton driver MW bunches
to be distributed over several targets [12]. The pion bunches produced at the targets
are focused by a horn system into a few meter long funnel, which merges them along a
common axis (Fig. 12.14). A straight pion decay/muon collect channel follows which
may be either a FODO line or a long solenoid. The main goal in designing a muon
collect channel is to maximize the pion and muon transmission. The overall length
of that funnel.+FODO collect channel depends on the muon momentum, .≈ 30 m
typically for parent pion beam momentum .≈ 200 ∼ 400 MeV/c (Fig. 12.15).
As an illustration, Fig. 12.15 shows sample parent pion and decay muon tra-
jectories, in the case of a funnel+FODO channel tuned on 300 MeV/c and with
initial pion bunch densities parabolic in x and y coordinates, total emittances
.ε x /π = ε y /π = 0.01 m, initial pion momentum distribution . p ∈ [200, 400] MeV/c,
uniform. Tracking trials in these hypotheses show that pion losses are marginal
along a .0.3 m aperture radius funnel channel, muon losses are marginal as well along
a .0.5 m aperture radius FODO line. The effect of beam line aperture is illustrated in
Fig. 12.16.
The latter result assumes that all pions and muons that make it to the end of the line
(s .= 30 m) are counted. In actuality, downstream lines (muon bunching, accelerators,
etc.) only allow so much admittance, accounted for this is illustrated in Fig. 12.17:
an admittance is defined, namely .εx /π = 0.01, 0.02, or .0.04 m, and a best matching
ellipse is computed, i.e., the surface of the ellipse .εx is imposed, beam parameters
.α x,y and .β x,y are free, they are obtained by matching with the sole criterion of best

transmission, i.e., greatest count within the ellipse boundary. As part of that matching
procedure, the FODO channel strength may be fine-tuned so that the matching ellipse
find itself on-axis at the end of the line. The vertical phase-space is subjected to a
similar treatment.
The muon beam longitudinal emittance (Fig. 12.13) may be subjected to a bound-
ary as well, to account for the longitudinal admittance of the downstream lines: the
surface .εl of the ellipse is imposed, its center and beam parameters .αl and .βl are free,
the constraint is maximum count through the ellipse boundary (Fig. 12.18).
490 12 Beam Lines

Fig. 12.14 Funnel section of the upstream part of a multiple-target pion and muon collect chan-
nel [12]. On the top synoptic the boxes on both sides of the axis represent the quadrupoles (focusing,
up, or defocusing, down), boxes straddling the axis represent bending dipoles. The pion production
target is at the left end, a long muon collect channel follows to the right (Fig. 12.15). This graph
shows the off axis trajectory of the beam centroid and the optical functions. A graph obtained using
the MAD program [14]

Fig. 12.15 Sample rays in a muon collect channel, including an upstream 9 m long funnel and a
downstream 32 m long FODO line. Left: parent pions. Right: decay muons. Trajectories end where
particles hit the vacuum pipe. The latter is 0.3 m in radius along the funnel, 0.5 m in radius along
the FODO line
12.2 Basic Concepts and Formulæ 491

Fig. 12.16 Transmission efficiency of the the pion, muon or pion+muon beams, versus distance
along a .π + μ collect channel (relative to the total number of parent pions, at s=0), for either 0.2,
0.3, 0.4 or 0.5 m radius collimation along the FODO line, and given an aperture radius .> 0.3 m
along the funnel. A 0.5 m aperture radius FODO line yields 85% transmission at s .= 30 m

Fig. 12.17 Horizontal phase

space at end of the 0.5 m
radius collimation FODO
channel (s .= 30 m). The
ellipses are the best match to
the .π + μ beam, with
emittances imposed, namely,
.εx /π = 0.01, 0.02 or
.0.04 m. The FODO channel
strength has been fine-tuned
to center the beam ellipses

Fig. 12.18 Time-energy

phase space at the end of the
FODO channel (s .= 30 m),
and best matching ellipse, for
the four different admittance
values .εl = 0.1, 0.5, 1 and
.2 π eV.s
492 12 Beam Lines

Fig. 12.19 Transmission

efficiency (relative to the
total number of parent pions,
at s .= 0) for various
transverse admittance at
s .= 30 m, as a function of the
longitudinal admittance .εl /π
at s .= 30 m

Results may be summarized as in Fig. 12.19 which shows the transmission effi-
ciency as a function of the longitudinal acceptance .εl /π at line end
(s .= 30 m), given .εx /π = ε y /π = 0.01 m, or .0.04 m, or infinite transverse accep-
tance. Parent pions have been launched with . p ∈ [200, 400] MeV/c, uniform, and
with .εx /π = ε y /π = 0.01 m. The FODO section of the AG channel has a 0.5 m
radius collimation, the beam line is tuned to 240 MeV/c for optimum transmission.

12.2.4 Low Energy Spin Rotator

In a polarized beam installation, prior to injection into downstream stages, a linac

for instance, polarization generally needs be oriented with respect to the beam prop-
agation axis, from longitudinal at the polarized electron source (Fig. 12.20) [15, 16].
In electron installations, Wien filters may be used as part of the polarization
rotation equipment.
An expression for the spin rotation over a distance . L may be obtained by analogy
between the Lorentz equation and the spin motion equation. Namely, from

dv B BL
. = v× ⇒ trajectory deviation = (12.24)
ds Bρ Bρ

wherein .B is the velocity precession vector, one infers

dS ω ωL
. = S× ⇒ θsp = (12.25)
ds Bρ Bρ

with .ω the spin precession vector. The spin angular momentum results from two
torques: from the magnetic field and from the electric field. The respective precession
vectors can be expressed under the form
12.2 Basic Concepts and Formulæ 493

Fig. 12.20 Low energy (.∼150 keV) beam line at CEBAF. The first Wien filter (VW, vertical)
downstream of the photo-guns rotates the polarization from longitudinal to vertical. The second
Wien filter (HW, horizontal) rotates the polarization in-plane to compensate precession of CEBAF
transport magnets. Solenoids (SS) ensure additional polarization rotation requirements [15]

Fig. 12.21 A sketch of the

reference frame used and
electron motion in pure . E
field (dots, red) or pure . B
field (dashes, blue). Under
the conjugate effect of both,
with . E Y = v B Z , the electron
trajectory is along the X axis
(green)

.ω B = (1 + a) B|| + (1 + aγ ) B⊥ (12.26)
( )
1 E× v
ωE = γ +a
γ +1 c2

with . B|| the projection of . B on the velocity direction and . B⊥ = B − B|| normal to
the latter. The anomalous magnetic moment for the electron is .a = 1.15965 × 10−3 .
Considering the hypotheses defined in Fig. 12.21, take . E || Y and . B || Z. The
condition for a straight electron trajectory is

. EY = v BZ (12.27)

In the present . E, B configuration, in particular . B ⊥ v always, one has

( )
1 E× v
ω = ω B + ω E = (1 + aγ ) B⊥ + γ
. +a (12.28)
γ +1 c2
494 12 Beam Lines

Fig. 12.22 Spin motion in the . B⊥ plane (the (X,Y) plane in Fig. 12.21) over the straight electron
trajectory, from longitudinal at s=0 to transverse at s .= 1.5 m

Taking . z unitary vector along Z, accounting in addition for . E ⊥ v, one has

. B = B Z z, E × v = −vE Y z = −v2 B Z z, ω = ω z (12.29)

With these ingredients .ω can be substituted in the .θsp expression above (Eq. 12.25)
to yield
( )
BZ L 1 BZ L
θ
. sp,th = (1 + aγ ) B−γ + a β2 = 30◦ (12.30)
Bρ γ +1 Bρ
from from E

This is illustrated in Fig. 12.22 which shows a .90◦ Y-rotation of the spin, from
. S || v to . S ⊥ v, along the straight trajectory of an electron through the Wien filter
. E, B field.

12.2.5 Synchrotron Radiation

Electron beams radiate in beam lines, which include the long beam lines which
recirculating linear accelerators are. This entails energy loss with effects on beam
dynamics, such as spiraling in bends and emittance growth. On the other hand, visible
radiation is used for beam diagnostics in circular accelerators, at high energy in the
case of hadron rings. The diagnostics system in these rings may be based on a main
12.2 Basic Concepts and Formulæ 495

magnet edge, on one or more dedicated short dipoles, or a short wiggler, or else, and
can be dealt with as a stand alone short beam line.
SR and its simulation in these beam line style of systems has been addressed
in earlier chapters. Topics addressed there are summarized below, together with
further additional material relevant to numerical raytracing simulations, possibly
documented in zgoubi sourceforge repository examples.

SR Energy Loss

Energy loss by synchrotron radiation has been introduced in Chap. 5, Sect. 5.2.3, as
its effects were first observed, and overcome, in a betatron. SR loss in a linear FFAG
beam line, and its effects on spin motion, are addressed in Chap. 11, Sect. 11.2.3.
First order perturbation methods [17] can be resorted to in order to validate raytrac-
ing outcomes in long beam lines, such as longitudinal and transverse beam matrix
perturbation, and transverse and longitudinal emittance growth.
The 1.2 km long final focus delivery system of the 250 GeV TESLA linear collider
test facility (TTF) for instance, has been the object of such cross-checks. An ad hoc
Tech. Note details the method and outcomes [18]. Zgoubi simulation files are also
available [19].

SR Poynting

Visible SR and its use for beam diagnostics have been introduced in Chap. 8,
Sect. 8.2.3, as the first observation happened at a weak focusing synchrotron. Several
publications are available which the reader can refer to, reporting zgoubi simula-
tions and/or analytical modeling of SR in short sections of rings. From the point
of view of the SR electric field impulse and it spectrum, these can be handled and
treated as short beam line segments. Published material includes:
– 45 GeV electron beam diagnostics and SR interference, at LEP in the 1990s, from
a series of four short dipoles, Sect. 9.2.6 and Refs. [20–22]
– diagnostics at the LHC, using visible SR from a dipole edge+wiggler string [23]
– 270 GeV proton beam diagnostics at the SPS, using edge radiation, Refs. [24, 25],
a case of edge observation similar to the two-dipole configuration of Exercise 9.6
– visible proton SR from an undulator at the SPS [26], a case which can be simulated
using the UNDULATOR element in zgoubi.
496 12 Beam Lines

12.3 Exercises

Note: some of the input data files for these simulations (the long files!) are available
in zgoubisourceforge repository at https://sourceforge.net/p/zgoubi/code/HEAD/
tree/branches/exemples/book/zgoubiMaterial/beamLines/

12.3.1 Beam Expander

This exercise concerns optical tests planned at the G4 line of GANIL experimental
areas (Fig. 12.23) [3], regarding beam uniformization for hadrontherapy applica-
tion [3].
The linear optics of the line is first established, including proper locations for
horizontal and vertical non-linear lenses. The latter are then powered to uniformize
the transverse beam density, in one or both planes. The exercise is concluded with
the exploration of a uniform ribbon sweeping technique.

Fig. 12.23 G4 beam line

enclosure in GANIL
experimental areas [28].
Beam goes from bottom left
to top right. The first four
quadrupoles in the present
simulation sequence
(Q21–Q24, Table 12.1) are
located in the upstream part
of the line, Q21 is next to the
shielding at the lower left
corner of the room
12.3 Exercises 497

12.1 Construct a Beam Expander Line

Solution 12.1.
Consider Fig. 12.24. The beam expander section of the line follows G4’s Q21–Q24
quadrupole section. The optics is set to create a vertical waist and large .βx at the
horizontal uniformization octupole, OH, and a horizontal waist and large .β y at the
vertical uniformization octupole, OV. The four quadrupoles downstream of OV are
tuned to provide the desired size of the square image at the target, including .αx =
α y = 0.
The optical parameters of the optical sequence are detailed in Table 12.1.
(a) Explain how the beam uniformization line operates.
(b) Produce a simulation input data file of this expander, from the data in
Table 12.1. QUADRUPO can be used to simulate the quadrupoles, or MULTIPOL
as well; the latter is used to simulate the octupoles; BEND can simulate the final
horizontal and vertical scanning dipoles. The octupole lenses and final dipoles are
assumed off, for the moment.
(c) Produce the transport coefficients of the line from quadrupole Q21 to TARGET,
to third order. OBJET[KOBJ .= 6.1] can be used to generate a proper object for 3rd
order matrix computation by MATRIX.

Fig. 12.24 A beam expander/beam uniformization line. Optical functions .βY and .β Z along the
line, including a scan of .β Z at target, from 10 to 700 m. The first four quadrupoles (boxes above or
under the axis) are part of GANIL G4 beam line. The following eight ones are part of the expander
line extension. The rightmost two scanning dipoles (boxes straddling the axis) are aimed at allowing
a sweeping of the uniformized beam spot across the target, located at .s = 15.9 m
498 12 Beam Lines

Table 12.1 G4 beam line and added expander section. Fields in quadrupoles are at 10 cm radius
Element Name Length (cm) Field (kG)
Final section of GANIL G4 beam line:
QUADRUPO Q21 30.4 –3.674189
DRIFT DR12 39.6
QUADRUPO Q22 30.4 2.067729
DRIFT DR23 187.5
QUADRUPO Q23 30.4 1
DRIFT DR34 39.6
QUADRUPO Q24 30.4 –2.791886
DRIFT SDP5 50
Additional tuning knobs:
QUADRUPO Q25 30.4 1
DRIFT SDP5 50
QUADRUPO Q26 30.4 –3
DRIFT SDP5-4 90
Expander/uniformization section:
MULTIPOL OH 10.000 TBD.(∗) Horizontal
octupole
DRIFT SDP5 50
QUADRUPO QV1 35.000 2.805048
DRIFT SD1 100
QUADRUPO QV 35.000 –3.034423
DRIFT SDP8 80
MULTIPOL OV 10.000 TBD.(∗) Vertical octupole
DRIFT SDP4 40
QUADRUPO QPL3 35.000 1.896253
DRIFT SDP5 50
QUADRUPO QPL2 35.000 2.464301
DRIFT SDP5 50
QUADRUPOQ QPL1 35.000 –0.1375607
DRIFT SDP5 50
QUADRUPO QPL0 35.000 –3.357716
DRIFT SD1LB 10
H. BEND BH 20 0 Scanning dipoles
V. BEND BV 20 0
DRIFT SD1LA 250
MARKER TARGET
(*) Field value to be determined, as part of the exercise
12.3 Exercises 499

12.2 Expander Line Optics

Solution 12.2.
The horizontal and vertical optical functions at entrance of Q21 in GANIL G4 line
are set to, respectively,

αY = −1.435, βY = 3.00933 m, εY /π = .05 × 10−6 m

(12.31)
α Z = −6.129, β Z = 44.3186 m, ε Z /π = 10−6 m
.

(a) Give a graph of a few tens of trajectories through the line, in the horizontal and
vertical planes, taking initial coordinates at random in a Gaussian distribution with
beam parameters given in Eq. 12.31. MCOBJET[KOBJ .= 3] can be used for that.
(b) Compute and plot the optical functions along the beam line, using the
.β Z (TARGET) = 10 m expander section settings of Table 12.1. Produce the beam
matrix at target.
Compute and plot the optical functions for the additional four settings,
.β Z = 100, 300, 500, and 700 m, Table 12.2. Check that .β Z values at TARGET are
as expected.

12.3 Transverse Uniformization; Beam Scan

Solution 12.3.
Consider the expander/uniformization section optical settings of Table 12.2: they
sample the quadrupole strengths (the vertical octupole strengths are to be deter-
mined as part of the exercise), necessary to achieve the rectangle beam spot sweep
at target displayed in Fig. 12.25. Note that the vertical octupole only is used, here,
uniformization is in the vertical direction only, the horizontal density remains that
defined by MCOBJET, i.e., Gaussian.
(a) Determine the OV octupole field values for a uniform vertical density at target,
in the five different .β Z cases.

Table 12.2 Strength in expander/uniformization section quadrupoles for five different values of.β Z
at target. A continuous sweep of the rectangular spot across the target is obtained by interpolation,
from these discrete values, of the field in the final lenses and in the BH and BV bends of the line
.βz at target (m)

10 100 300 500 700

QV m.−2 .−3.0342 .−3.2068 .−3.2996 .−3.3529 .−3.3933

OV octupole m.−4 Strengths to be determined

QPL3 m.−2 1.8962 2.884847 3.207578 3.230446 3.211739
QPL2 m.−2 2.4643 2.3276 2.0705 2.1048 2.1589
QPL1 m.−2 .−0.13756 .−1.4562 .−1.1073 .−1.1073 .−1.1073

QPL0 m.−2 .−3.3577 .−1.1889 .−1.0442 .−0.95639 .−0.91447

500 12 Beam Lines

Fig. 12.25 Simulation of a continuous rectangle sweep. Twenty steps are shown here. The rectangle
beam density is uniform in the vertical direction, Gaussian in the horizontal direction

Determine the dipole field values in the scanning dipoles:

– in the case of BH, at spots number 1 (rightmost), 6 and 20 (leftmost),

– in the case of BV, at spots number 1 (rightmost), 6, 12 and 20 (leftmost),

proper to position the rectangular spot as displayed in Fig. 12.25, at the five
β Z (TARGET) values.
.
Reproduce Fig. 12.25, using REBELOTE[NRBLT.= 19] to iterate, and SCALING
to vary the expander/uniformization quadrupole settings, the BH and BV fields, and
the vertical octupole field, over the 20 iterations.
(b) Using the results of question 12.3(a), produce a 2D-uniformized .6 × 6 cm2
beam spot at target. Give a graph of the transverse cross-section.
(c) Reproduce some of the transverse beam density patterns downstream of the
octupoles, in a similar way to Fig. 12.5. HISTO[PRINT] can be used, it prints out
its histograms to zgoubi.res, and to zgoubi.HISTO.out whose content can then be
plotted (using gnuplot, for instance).

12.4 Animation: Transverse Scan by a Uniform Beam

Solution 12.4
Produce an animation of a transverse scan by a 1,000-particle rectangle with uniform
vertical density.
12.3 Exercises 501

12.3.2 Nano-Probe .E, B Final Doublet Achromat

12.5 Construct a Final Focus Doublet

Solution 12.5.
(a) Construct a beam optics model of the optical sequence sketched in Fig. 12.26
leaving first the electrostatic component zero, with B set to provide the image distance
indicated in the figure. The field-free sections can use DRIFT, the two quadrupoles can
be simulated using EBMULT. Note that, as long as the E (respectively B) component
is zero, QUADRUPO (respectively ELMULT) can be used instead.
Produce the first and second order transport matrices of this system from paraxial
trajectory tracking, using MATRIX, together with OBJET[KOBJ .= 6] to define a
proper set of paraxial rays. Check the second order geometric coefficients: they are
expected to be null [29]. Check the second order chromatic coefficients: the non-
coupled ones are expected to be non-zero [29].
Plot the horizontal Y(s) and vertical Z(s) projections of the paraxial trajectories
used to compute the transport coefficients.
(b) Assume a point object with 0.2 mrad rms divergence, Gaussian, in both planes,
and take.10−3 rms momentum spread, uniform. MCOBJET[KOBJ .= 1] can be used
to define this initial coordinate distribution.
Check the location of the vertical and horizontal waists, in the image space.
IMAGE and IMAGEZ can be used for that.
Provide the histograms of these initial random distributions, using HISTO, pos-
sibly HISTO[PRINT].
Produce a plot of the beam cross-section, of the horizontal and vertical phase
spaces, and of the projected coordinate histograms, downstream of the final 25 cm
drift plane.
(c) Repeat (a) and (b) using an electrostatic doublet instead. This requires using
ELMULT or EBMULT.

QF QD
OBJET

IMAGE
4.9

559 10.2 10.2 25

Fig. 12.26 Principle scheme of a QF-QD (focusing-defocusing) final focus second-order achromat
for a nano-probe ion beam. The scheme includes design lengths and distances used in the exercises:
559 cm from object to entrance of first quadrupole, 4.9 cm between quadrupoles, 25 cm from exit
of second quadrupole to image, and 10.2 cm quadrupole length
502 12 Beam Lines

12.6 E, B Lens, Hard-Edge Model

Solution 12.6.
(a) Given Exercise 12.5 conditions, which assume hard-edge modeling of the two
lenses, now switch on the electric component (this imposes using EBMULT), setting
.φ and . B to fulfill the following two constraints:
(i) preserve the focal distance,
(ii) ensure cancellation of second order chromatic aberrations.
Produce the second order transport matrix, compare with the previous E=0 case.
(b) Track 5,000 particles through the achromat. Assume a point object with
0.2 mrad rms divergence, Gaussian, in both planes, and take.10−3 rms momentum
spread, uniform.
Produce the beam cross-section (point-spread function), the horizontal and verti-
cal phase spaces, and the projected coordinate histograms, downstream of the final
25 cm drift. Compare with the conditions of Exercise 12.5. FAISTORE or FAISCNL
can be used to store particle data at that location.

12.7 E, B Lens With Fringe Fields

Solution 12.7.
Add fringe fields, considering an unrealistically large radius at pole tip of 5 cm (for
the sake of enhanced possible adverse effects). The same geometry is maintained
(Fig. 12.26).
(a) Find the new values of the E and B components of the lenses, for a paraxial
image at the same location together with cancelled chromatic coefficients. FIT can
be used for that matching.
Produce the second order transport matrix, compare with the hard-edge case.
(b) Plot some of the . E(s) and . B(s) field components experienced along a straight
line across the doublet, e.g., .Y = 1 cm or . Z = 1 cm. EBMULT[IL .= 2] can be used
to store particle data during stepwise raytracing across the lens. Plot some of the first
and second order field derivatives as well.

12.8 Animation: Dynamical Squeeze of Point-Spread Function

Solution 12.8.
Build the following graphic animation: a dynamic squeeze of the image by slowly
switching on the E, B parameters, while preserving the focal distance of the final
quadrupole doublet. REBELOTE[K .= 99, IOPT .= 1] can be used to repeat transport
through the optical sequence with EBMULT parameters varied, pass after pass.
12.3 Exercises 503

12.3.3 .π → μ + ν In-Flight Decay, Bunch Densities

In a first part in the following simulation exercises, the evolution of pion and muon
beam distributions resulting from in-flight pion decay

π →μ+ν
.

is investigated. A Monte Carlo method is used to simulate the process. Unless oth-
erwise specified, a parent bunch comprised of . N0 = 104 to .106 pions is considered.
The bunch is launched at .s = 0 with zero transverse emittances, zero length, and
. pπ ∈ [100, 500] MeV/c (either a set of initial discrete . pπ values, or some random

distribution). MCDESINT defines the daughter particle and switches on the decay
process. PARTICUL is needed prior, in order to define the parent particle.
In a second part, a muon collect beam line is simulated [11–13]. Its optical prop-
erties and collect efficiency are investigated.

12.9 Parent Pion Bunch

Solution 12.9.
(a) Figure 12.9 shows the theoretical pion momentum density at various distances
along the muon collect channel (Eq. 12.14), assuming initial random uniform . pπ ∈
[100, 500] MeV/c. Produce these distributions from Monte Carlo simulations.
MCOBJET[KOBJ .= 3] can be used to generate the initial bunch. A series
of successive DRIFTs allows to transport the beam over distance increments;
HISTO[PRINT] can be used, histogram plots can be obtained reading from
zgoubi.HISTO.out which the PRINT command produces. An other possibility is
to use DRIFT[split, IL .= 2] and plot the histograms from particle data so logged in
zgoubi.plt.
(b) Produce the distance dependence of the pion bunch average momentum
(Fig. 12.10). This can use DRIFT[split, IL .= 2]

12.10 Muon Bunch

Solution 12.10.
(a) Figure 12.11 shows muon momentum density plots (Eq. 12.19), for three different
initial discrete pion momenta values . pπ = 100, 250, 500 MeV/c. Produce these
distributions from Monte Carlo simulations.
(b) Figure 12.12 shows the theoretical muon momentum density at various dis-
tances along the muon collect channel (Eq. 12.21). Reproduce these distributions
from Monte Carlo simulations.
(c) Produce the distance dependent average momentum of the muon bunch
(Fig. 12.10).
(d) Reproduce the final longitudinal muon bunch phase space of Fig. 12.13. FAI-
STORE can be used to store particle data at the end of the line.
504 12 Beam Lines

12.3.4 Pion Funnel and Muon FODO Collect Channel

The Neutrino Factory muon collect channel of section “Neutrino Factory Muon
Collect Channel” is considered here. The funnel section starts at s .= 0 (downstream
of the pion production target and focusing horn system [12], not concerned, here).
The orbit excursion and betatron functions along are displayed in Fig. 12.14. In this
exercise, the funnel is followed by a 14-cell, .≈ 20 m long FODO channel.
The beam line sequence and nominal parameters are detailed in Table 12.3. Initial
values of the optical functions are given in Table 12.4. Using ray-tracing methods
allows accurate modeling of magnetic fields in the large aperture optical elements
traversed by the beam (up to 1 m radius for some of the upstream ones in the funnel),
and on the other hand allows detailed simulation of on-flight .π → μ decay based on
stepwise Monte Carlo.

Table 12.3 Design parameters of the funnel.+FODO collect channel. Field values assume a refer-
ence pion momentum p .= 300 MeV/c. . B0 is the dipole component in combined function dipoles.
The quadrupole field . B1 is at 10 cm radius
Element Name Length (cm) . B0 (kG) . B1 (kG)

Funnel
DRIFT DRF1 50
MULTIPOL DQ1 100 .−0.838529167 0.291349611
DRIFT DRF2 40
MULTIPOL Q2 100 0 .−0.617467029

DRIFT DRF3 65
MULTIPOL Q3 100 .−0.250226990 0.81443677
DRIFT DRF2 40
MULTIPOL Q4 100 0 .−0.73629954

DRIFT DRF4 40
MULTIPOL DIP 40 0 .−1.35257476

DRIFT DRFM1 44
MULTIPOL QM1 40 0 1.79773547
DRIFT DRFM2 107
MULTIPOL QM2 40 0 .−2.40617729

DRIFT DRFM3 40
FODO cell
DRIFT DRF 18.66435
MULTIPOL QF 40. 0 3.00591
DRIFT DRF 37.3287
MULTIPOL QD 40. 0 .−3.00591

DRIFT DRF 18.66435

12.3 Exercises 505

Table 12.4 Reference trajectory coordinates and optical functions at funnel entrance (s .= 0)
'
. x co /x co m/rad 0.3943/0.09036
. D x /D x
' m/rad 0/0
.βx /βz m 4/4
.αx /αz 0/0

12.11 Beam Line Simulation

Solution 12.11.
Refer to the design parameters given in Table 12.3.
(a) Simulate this beam line, produce a graph of the reference trajectory. Produce
the transport matrix.
A proper 13-particle object for MATRIX computation can be defined using
OBJET[KOBJ .= 5]. The simulation of straight sections uses DRIFT, dipoles and
combined function dipoles in the funnel can use MULTIPOL. Quadrupoles can use
QUADRUPO, or MULTIPOL (the latter would allow introducing multipole defects,
if desired).
(b) Optical functions at s .= 0 are given in Table 12.4, produce a graph of the
optical functions along the line.
Give their values at end of funnel and at end of FODO channel.
OPTICS can be used to transport the optical functions, with initial values defined
using OBJET[KOBJ .= 5.1].
(c) Produce a graph of paraxial horizontal and vertical phase space invariants
at start and end of the channel, at 250, 300 and 400 MeV/c, given initial elliptical
invariant values .εx = ε y = 2.5 π mm at the entrance of the funnel.

12.12 Bunch Densities

Solution 12.12.
A .104 pion bunch is considered, launched at entrance of the funnel with beam and
emittance parameters as defined in Table 12.4. The bunch can be defined using the
random coordinate generator MCOBJET[KOBJ .= 3].
In-flight decay requires additional keywords: it requires specifying the parent par-
ticle, pions, this is done using PARTICUL, as well as the daughter particle, muon,
this is part of the data of the Monte Carlo in-flight decay switch MCDESINT. Col-
limation is introduced to stop decay muons which reach the channel aperture, long
collimation can use CHAMBR, local collimators can use COLLIMA.
(a) Plot the transverse coordinates of a few particles, along the line, in a similar
way to Fig. 12.15.
(b) Extend the channel to 100 m distance, take a 0.5 m radius collimation all the
way. Take .εx /π = ε y /π = 0.01 cm initial pion beam emittances, and
. p ∈ [200, 400] MeV/c uniform.
506 12 Beam Lines

Produce a graph of the average momentum of respectively.π ,.μ and.π + μ beams.

Plot transverse pion bunch densities at start, 20 and 50 m in a similar way to
Fig. 12.9, and transverse muon bunch densities at 100 m in a similar way to Fig. 12.12.
Check qualitatively by comparison with these theoretical densities.
(c) For four different longitudinal ellipse surfaces: .εl = 0.1, 0.5, 1 and .2 π eV s,
determine the phase space ellipse parameters (centering and shape) which correspond
to a maximum acceptance at the end of the collect channel, in a similar way to
Fig. 12.18. Phase space collimation with COLLIMA[IFORM .= 11–16] can be used
for that.
(d) Indicate how the FIT procedure can be used to optimize the collect channel
design parameters for maximized transmission.

12.3.5 Low Energy Spin Rotator

This exercise series addresses electron spin dynamics in a spin rotator based on a
Wien filter. Stepwise ray-tracing techniques allow detailed insight in the dynamics
of spin motion through that . E, B device.
The analytical modeling WIENFILTER is used in these exercises (TOSCA could
be used instead, would a . E, B field map be available). . E and . B fields are first taken
hard-edge so to allow tight comparison with theory, fringe field are added, next, their
effect is assessed and compensated.
The reference frame in the exercises is that of Fig. 12.21, i.e., longitudinal axis X,
horizontal and vertical transverse axes Y and Z.
The WIENFILTER considered is comprised of three 0.5 m long segments, rotating
the spin by .30◦ each. A hard edge field model is considered first, then fringe fields,
which have a noticeable effect, are added.

12.13 Spin Rotation, Spin Matrix

Solution 12.13.
A hard-edge field model WIENFILTER segment is considered here.
(a) Calculate from theory, and check from raytracing, the value of the electric and
magnetic field components in a 0.5 m long Wien filter, to obtain a straight electron
trajectory and .30◦ spin rotation. FIT can be used to obtain the required E and B
values.
(b) Produce the spin rotation matrix. OBJET[KOBJ .= 2, IMAX .= 3] and SPN-
PRT[MATRIX] can be used for that.
12.3 Exercises 507

12.14 Integration Step Size

Solution 12.14.
The effect of integration step size on the convergence of the numerical integration is
addressed in this exercise. In both simulations below, the do-loop REBELOTE[IOPT
.= 1] can be used to repeat spin transport through the optical sequence, concurrently
with changing the step size.
(a) Produce a graph of the variation of particle position and angle at the exit of
the Wien filter, as well as spin rotation angle, as a function of integration step size,
with . E Y and . B Z maintained constant (at their theoretical values).
(b) Produce a graph of the required variation of . E Y and . B Z (found using FIT)
to recover straight trajectory and .30◦ spin rotation, as a function of integration step
size.

12.15 Add Fringe Fields

Solution 12.15.
To simulate fringe fields, use Enge fringe field coefficient values

.C0 = 0.2401, C1 = 1.8639, C2 = −0.5572, C3 = 0.3904, C4 = C5 = 0

(a) First, take .λ E = λ B = 5 cm, at entrance and exit of a 50 cm segment.

Determine a proper value of the numerical integration extents . X E and . X S , at both
ends.
Determine an integration step size.
(b) .λ E and .λ B are in general different, owing in particular to different gap heights.
A consequence is that the theoretical field values, and in particular the condition
. E Y = v B Z , no longer ensure a straight trajectory.
Compute the dependence of the .δ E Y /E Y and .δ B Z /B Z field adjustments, proper
to recover particle trajectory and .30◦ spin rotation, on the field fall-off extent ratio
.λ B /λ E . FIT can be used to find . E Y and . B Z , and REBELOTE can be used to repeat
the sequence with prior change of .λ B .
Produce a graph of sample reference trajectories, Y(X) along the Wien filter, for
various values of .λ B /λ E . Check the spin rotation angle, as per the FIT procedure.

12.16 Transport of a 6D Polarized Bunch

Solution 12.16.
Launch through the Wien filter a .104 electron bunch with normalized rms emit-
tances.βγ εY = βγ ε Z = 10π µm, Gaussian transverse densities truncated at 4 sigma,
momentum spread uniform over.[−10−4 , +10−4 ]. MCOBJET[KOBJ=3] can be used
to generate this bunch.
Produce sample trajectories and . E Y (s) and . B Z (s) fields, across the Wien filter.
Produce the phase spaces and projected densities, spin rotation and densities, at
the downstream end of the complete, 3-segment, 1.5 m Wien filter.
508 12 Beam Lines

12.3.6 Synchrotron Radiation

12.17 Synchrotron Radiation in a Long FFAG Beam Line

See “eRHIC 22 GeV ERL. Synchrotron Radiation. Polarization”, a linear FFAG beam
line, Exercise 11.2.

12.4 Solutions of Exercises of This Chapter: Beam Lines

12.4.1 Beam Expander

12.1 Construct a Beam Expander Line

(a) Expander line operation.
The way the expander/uniformizer line described in Table 12.1 operates is as follows
(see Sect. 12.2.1).
• At the downstream end of the line (TARGET):
– The beam has zero divergence, its transverse dimensions are a few millimeters
to centimeters (Fig. 12.25). Four quadrupoles, QPL3-0, are tuned to adjust four
variables: .αY,Z , βY,Z .
– A final 3 m long section, SD1LB-SD1LA, includes two scanning dipoles, hori-
zontal (BH) and vertical (BV).
• Beam uniformization at the target:
– one octupole is needed per transverse dimension, they are tuned individually;
– the horizontal (respectively vertical) uniformization octupole is located at a
vertical beam waist in a region of large .βY (respectively, horizontal waist and
large .β Z ).
• Upstream end of the line: it includes the downstream end of GANIL G4 line, its
role is to match GANIL beam to the downstream expander/uniformizer section.
Six quadrupoles are available for that, the G4 quadrupoles Q21–Q24 and two
additional ones, Q25 and Q26 for optics setting flexibility.

(b) A simulation of the optical sequence

The optical sequence resulting from Table 12.1 data is given in Table 12.5. It starts
with OBJET[KOBJ .= 6.1] which creates a 102 particle sample, for computation of
the transport coefficients of the line up to 3rd order, by MATRIX. The latter can be
found at the end of the sequence. The reference rigidity is set to . Bρref ≡ B O R O =
1000 kG cm, as the actual rigidity of the beam does not matter here. One advantage
in taking . B O R O = 1000 kG cm is that, with . R0 = 10 cm the reference radius of the
quadrupoles, the field at the pole identifies with the strength of the element:
12.4 Solutions of Exercises of This Chapter: Beam Lines 509

Bpole [kG] / R0 [cm]

. Strength K [m−2 ] =
Bρ[kG cm]

(c) Transport coefficients.

The transport coefficients can be found in zgoubi.res execution listing, under
MATRIX at the bottom of the file. An excerpt is given in Table 12.6. Note that trans-
port coefficients up to second order can be obtained with OBJET[KOBJ .= 6], which
generates an ad hoc 61 particle sample for computation by MATRIX. OBJET[KOBJ
.= 6.1] generates a 102 particle sample for the computation of transport coefficients

up to third order.

12.2 Expander Line Optics

(a) A few trajectories, from start to end.

The simulation input data file of Table 12.5 is used to raytrace and plot particle
coordinates along the line. The only change needed is to replace OBJET by the
following MCOBJET:

’MCOBJET’
1000. ! Reference rigidity.
3
80 ! 80 particles defined here.
2 2 2 2 2 2 ! All six coordinates are sorted at random in a Gaussian density.
0. 0. 0. 0. 0. 1. ! Reference coordinates, at start of the structure, including D=1.
-1.435 3.00933 .05E-6 3 ! alpha_Y; beta_Y; epsilon_Y/pi; cut-off.
-6.129 44.3186 1.E-6 3 ! alpha_Z; beta_Z; epsilon_Z/pi; cut-off.
0. 1. 1.E-96 3 ! alpha_l; beta_l; epsilon_l/pi; cut-off.
323456 134567 545679 ! Three seeds used by the random number generator.

In order to have particle coordinates logged in zgoubi.plt during raytracing, IL

is set to 2 in all lenses using the global command OPTIONS[IOPT .= 1, NBOP=1,
.plt IL.= 2]. For that very reason, a split command has been added in some of the
drifts using DRIFT[split, IL .= 2]. Graphs of 80 trajectories through the line, in the
horizontal and vertical planes, are given in Fig. 12.27.
510 12 Beam Lines

Table 12.5 Simulation input data file: expander/uniformization beam line G4_beamLine.inc. Uni-
formization octupoles and BENDs are off at this stage. This input data file defines two seg-
ments (using LABEL1’s): S_G4 to OH_dwn on the one hand, S_G4 to Target on the other
hand, for use in INCLUDEs in subsequent exercises. Note this file is available in zgoubi
sourceforge repository at https://sourceforge.net/p/zgoubi/code/HEAD/tree/branches/exemples/
book/zgoubiMaterial/beamLines/expander/MATRIX_O3/
12.4 Solutions of Exercises of This Chapter: Beam Lines 511

Table 12.6 Outcomes of the simulation file of Table 12.5. These are excerpts of zgoubi.res output
file, truncated for shortness

(b) Optical functions.

Different methods can be resorted to, to compute the optical functions. Two possi-
bilities are detailed in the following.
- first method
OPTICS[all, PRINT] allows transporting initial values of the optical functions.
OPTICS works in a similar way to TWISS, except for the fact that TWISS assumes
the sequence to be a periodic structure and as a consequence computes the peri-
odic functions at the start of the sequence (using Eq. 14.49) whereas OPTICS does
512 12 Beam Lines

Fig. 12.27 Trajectories through the expander beam line. Left: horizontal plane; right: vertical plane.
Note the different vertical scales. The horizontal uniformization octupole (OH) is at a location of
large horizontal excursion, small vertical excursion, the reverse holds for the vertical octupole (OV).
A graph obtained using zpop: menu 7; 1/1 to open zgoubi.plt; 2/[6, 2] to select.Y versus distance (or
2/[6, 4] to select . Z ); 7 to plot. Option 3/10/10 would plot trajectories as continuous lines, whereas
the present default 3/10/9 plots dots, does not connect the stepwise data

not make that assumption and thus requires these to be provided. OBJET[KOBJ .=
5.1] is used to generate a 13 particle sample, proper for MATRIX computation, as
OPTICS uses the same Fortran routines as MATRIX to compute transport matrices.
OBJET[KOBJ .= 5.1] also allows providing the initial values of the optical func-
tions. OPTICS[all, PRINT] causes computation of optical functions after (almost)
all keywords in the optical sequence; the PRINT argument causes these to be logged
in zgoubi.OPTICS.out (in a similar way that TWISS causes the optical functions to
be logged in zgoubi.TWISS.out). The input data file for this simulation is given in
Table 12.7.
A gnuplot script file found in zgoubi toolbox, gnuplot_OPTICS.gnu [31], can
be used to plot zgoubi.OPTICS.out content; this script is called using a SYSTEM
command (Table 12.7). OPTICS computation is also an opportunity to check

– the value of the reference orbit, with respect to which transport coefficients and
the resulting optical functions are computed,
– the degree of non-symplecticity of the numerical integration of the motion, via the
first order mapping determinants, for instance.

The graphs resulting from the execution of Table 12.7 simulation file, via the
SYSTEM command, are displayed in Fig. 12.28.
- second method
The input simulation file is the same as in the first method, Table 12.7. Set
OPTIONS[.plt, IL .= 2], uncomment (in case it is commented) the split command
DRIFT[split, IL.= 2] in various drifts where it is present, this will cause stepwise data
to be logged in zgoubi.plt. This allows using the post-treatment tool betaFromPlt,
found in zgoubi toolbox [32], to compute optical functions from the stepwise data
read in zgoubi.plt. The second gnuplot script under SYSTEM, Table 12.7, found
as well in zgoubi toolbox [32], performs the two tasks, first a system call to
12.4 Solutions of Exercises of This Chapter: Beam Lines 513

Table 12.7 Simulation input data file: computation of the optical functions along the
expander/uniformization beam line, using OPTICS. This input data file INCLUDEs the segment
from the G4_beamLine.inc file of Table 12.5 The first gnuplot script under SYSTEM plots the
optical functions as read from zgoubi.OPTICS.out (Fig. 12.28); the second gnuplot script (found
in zgoubi toolbox) resorts to betaFromPlt to transport the optical functions, and plots the latter
(Fig. 12.29)

Fig. 12.28 Left: optical functions along the expander line, case of .β Z = 10 m at target. Right:
the first order mapping determinants only weakly differ from one—note that a contribution to the
difference to 1 is the limited accuracy of the interpolation method [30, cf. MATRIX] that computes
the transport coefficients from the coordinates of the 13 particle sample
514 12 Beam Lines

Fig. 12.29 Optical functions along the expander line. Stepwise particle coordinates across opti-
cal elements are read in zgoubi.plt, they allow computing the transport matrix .[Ti j ](s ← s0 )
from the origin .s0 to .s, which in turn allows transporting the beam matrix .σ (s) = [Ti j ](s ←
s0 )σ (s0 )[T˜i j ](s ← s0 )

betaFromPlt, and then plotting the content of the resulting betaFromPlt.out. The
resulting graph is displayed in Fig. 12.29.
Beam matrix at target
This is an outcome of OPTICS[all]: this releases beam matrices at all optical elements
along the beam line. The beam matrix at target is found at the bottom of zgoubi.res
execution listing:

37 Keyword, label(s) : MARKER Target

Reference particle (# 1), path length : 1589.1000 cm relative momentum : 1.

BEAM MATRIX (beta/alpha/alpha/gamma), FINAL

3.00026 -2.153829E-05 0.00000 0.00000
-2.153829E-05 0.333284 0.00000 0.00000
0.00000 0.00000 9.99788 9.805930E-05
0.00000 0.00000 9.805930E-05 0.100014

The vertical betatron function takes the value .β Z = 9.997 m at the target, this
determines the vertical size of the rectangle in the presence of the vertical octupole
(Eq. 12.2). It can be observed that .αY = α Z = 0, beam ellipses for the linear setting
(octupoles off) are upright.
12.4 Solutions of Exercises of This Chapter: Beam Lines 515

Fig. 12.30 Optical functions along the expander line, for.β Z = 100, 300, 500 and.700 m at target,
using zgoubi.plt stepwise raytracing data to compute the beam matrix transport .σ (s) = T (s ←
s0 )σ (s0 )T̃ (s ← s0 )

Optical functions for .β Z = 100, 300, 500 and .700 m at target

They are produced as above, using the simulation input file of Table 12.7. The field
values in the quadrupoles of the expander/uniformization section (Table 12.5) are
given in Table 12.2. The simulation needs be repeated for each value of .β Z at target.
Results are given in Fig. 12.30.
Beam matrices so obtained from OPTICS[all, PRINT] are logged in
zgoubi.OPTICS.out, their values at TARGET can be found at the bottom of the
file, they can be found as well at the bottom of zgoubi.res execution listing, and
checked against the expected values, namely,

α = α Z = 0; βY = 3 m in all cases; β Z = 100, 300, 500 and 700 m

. Y

12.3 Transverse Uniformization; Beam Scan

The beam at target retains its Gaussian horizontal density as the horizontal
octupole is off, with rms width .σx = (βx εx /π )1/2 ≈ (3 × 0.05 × 10−6 )1/2 =
0.4 × 10−3 m ; the vertical beam profile is uniformized due to the vertical octupole,
and with total extent .∆Z in millimeter to several centimeter range depending on .βz
(Eq. 12.4).
516 12 Beam Lines

Table 12.8 In complement to Table 12.2: octupole strengths for beam uniformization
.βz at target (m)

10 100 300 500 700

Octupole m.−4 0.2 .−3.8 .−2.49539 .−2.8696 .−4.1106

(a) Vertical octupole field value.

The OV octupole integrated strength required to obtain a uniform vertical density at
target is determined using

1 cos3 ∆φ
. K 3 L|OV =
12εz βl2 sin ∆φ

a simplified form of Eq. 12.7. The values of the betatron amplitude .βl at the octupole
lens, and of the vertical phase advance .∆φ from the octupole to the target, in the
five different cases of .β Z value at target (i.e., five different cases of spot size), are
taken from the optical function values, as found in zgoubi.OPTICS.out or zgoubi.res
files produced in the question (d). The field values so determined can be found under
MULTIPOL[LABEL1 .= OV], they are recapped in Table 12.8; they appear under
SCALING[MULTIPOL OV] in the simulation input data file, Table 12.9.
Field values in the scanning dipoles (Table 12.5) are determined from the geo-
metrical data, Fig. 12.25. These values are found under BEND/BH and BEND/BV,
under SCALING keyword.
The input data file for this sweep simulation is given in Table 12.9. The resulting
graph of the sweep is displayed in Fig. 12.25.
(b) Two-dimensional uniform expansion.
A proper setting of QV and QPL3 to QPL0 lenses has to be re-computed, to achieve a
6 × 6 cm2 square beam cross-section. In particular it has to satisfy Eq. 12.4 for both
.

horizontal and vertical planes. The octupole strengths are derived from Eq. 12.7. The
field values so obtained can be found under SCALING in the simulation data file
(Table 12.10).
(c) Histograms along the line
Graphs of transverse beam densities along the line are obtained using
HISTO[PRINT], which causes a log of particle density histograms to
zgoubi.HISTO.out. Three HISTO have been inserted in the simulation data file
(Table 12.10), their outcomes are plotted using gnuplot, as part of the simulation file
input (SYSTEM keyword), Fig. 12.32. These histograms are logged in zgoubi.res
execution listing as well, Table 12.11.
12.4 Solutions of Exercises of This Chapter: Beam Lines 517

Table 12.9 Simulation input data file: a rectangle beam spot sweep at target (Fig. 12.25). SCAL-
ING is introduced to allow interpolating the field in magnets from pass 1 to pass 20, from field
values at five different timings. The number of passes is specified as REBELOTE[NRBLT .= 19].
The INCLUDed segment [S_G4:OH_dwn] is grabbed from G4_beamLine.inc file (Table 12.5).
Fields in all quadrupoles from QV on, as well as the B3 (octupole) component in MULTI-
POL[LABEL1 .= OV], are set to 1, and scaled to proper value using SCALING: this allows
varying these values at each one of the 19 passes, for proper image size and uniformized verti-
cal density. Note this file, and the gnuplot script for the animation in Exercise 12.4, are available
in zgoubi sourceforge repository at https://sourceforge.net/p/zgoubi/code/HEAD/tree/branches/
exemples/book/zgoubiMaterial/beamLines/expander/rectangleSweep/
518 12 Beam Lines

Table 12.10 Simulation input data file: uniform 2D beam cross-section at target (Fig. 12.31).
SCALING is used to set the proper field values in the various magnets. The INCLUDed segment
[S_G4:OH_dwn] is grabbed for G4_beamLine.inc file (Table 12.5). The HISTO[PRINT] commands
cause a log of particle density histograms to zgoubi.HISTO.out
12.4 Solutions of Exercises of This Chapter: Beam Lines 519

Fig. 12.31 Two-dimensional uniformized beam cross-section at target, and projected histograms

Fig. 12.32 Horizontal beam density histograms at entrance of QPL1 (1), entrance (2) and exit (3)
of QPL0

12.4 Animation: Transverse Scan by a Uniform Beam

An animation of the frozen rectangle sweep in Fig. 12.25 is obtained by plotting the
content of the zgoubi.fai file produced by running the sweep simulation of Table 12.9.
A possible gnuplot script for that is:

ip1=1; ip2=20 ; set xrange [-4:4] ; set yrange [-7:7]

do for [ip=ip1:ip2] {
plot "zgoubi.fai" u ($38<=ip ? $10 :1/0):($12) w p
} ; pause 4
520 12 Beam Lines

Table 12.11 Outcome of the HISTO keyword, in zgoubi.res execution listing. The histogram at
exit of QPL0, here

12.4.2 Nano-Probe .E, B Final Doublet Achromat

12.5 Construct a Final Focus Doublet

(a) Optical sequence. Trajectories and transport matrix.

The simulation input data file is given in Table 12.12, and follows the scheme given in
Fig. 12.26. Note the use of KOBJ .= 6 in OBJET, which creates 61 rays, so allowing
computation of transport coefficients to second order in the coordinates, by MATRIX
(using KOBJ .= 6.1 instead would create 102 rays and allow additional computation
of the 3rd order transport coefficients).

Out of zgoubi.res execution listing, the resulting first order transport matrix .[Ri j ]
at the image is the following:

-0.142195 4.125875E-07 0.00000 0.00000 0.00000 0.00000

-1.63380 -7.03261 0.00000 0.00000 0.00000 0.00000
0.00000 0.00000 -2.986608E-02 6.239409E-08 0.00000 0.00000
0.00000 0.00000 -5.94705 -33.4828 0.00000 0.00000
0.00000 0.00000 0.00000 0.00000 1.00000 0.00000
0.00000 0.00000 0.00000 0.00000 0.00000 1.00000
12.4 Solutions of Exercises of This Chapter: Beam Lines 521

Table 12.12 Simulation input data file: final doublet achromat. The FIT procedure is used to adjust
the field values in both quadrupoles in order to ensure 25 cm image distance. Once FIT is done, the
execution pointer proceeds in sequence, to MATRIX, DRIFT, and whatever may follow. Note this
file is available in zgoubi sourceforge repository at https://sourceforge.net/p/zgoubi/code/HEAD/
tree/branches/exemples/book/zgoubiMaterial/beamLines/nanoProbeAchromat/animation/
522 12 Beam Lines

Fig. 12.33 Projected coordinates Y(X) and Z(X), across the optical system from object to 20 cm
downstream of the image plane, of the 61 paraxial rays used to compute the transport coefficients
to 2nd order (X is the distance along the optical axis). The vertical bars materialize the quadrupoles
and (rightmost bar) the location of the paraxial image plane

It can be verified that . R12 ≈ 0 and . R34 ≈ 0, conditions for an image located 25 cm
downstream of the quadrupole doublet. These two constraints are realized by a FIT
procedure (Table 12.12), by varying the magnetic field in the quadrupoles.
The second order transverse geometric coefficients, under MATRIX in zgoubi.res
execution listing, appear to all be null, as expected. The non-coupled chromatic coef-
ficients .T1−2,1−2,6 , .T3−4,3−4,6 , and .T1−4,6,6 are non-zero, namely, out of zgoubi.res:

1 16 1.15 1 26 6.27 1 36 0.00 1 46 0.00 1 56 0.00 1 66 0.00

2 16 2.85 2 26 15.4 2 36 0.00 2 46 0.00 2 56 0.00 2 66 0.00
3 16 0.00 3 26 0.00 3 36 1.02 3 46 5.89 3 56 0.00 3 66 0.00
4 16 0.00 4 26 0.00 4 36 4.72 4 46 27.0 4 56 0.00 4 66 0.00

Figure 12.33 displays the projected coordinates of the 61 paraxial rays used to
compute the transport coefficients to 2nd order, across the optical system. The gnuplot
script used for that is given in Table 12.13.

(b) Optical imaging. Locations of waist.

Table 12.14 gives the input data file to generate the required random object and
perform the tracking.
The gnuplot script given in Table 12.15 is used to produce Fig. 12.34, which shows
the initial horizontal and vertical Gaussian angle distributions, and the momentum
uniform distribution, resulting from MCOBJET keyword.
12.4 Solutions of Exercises of This Chapter: Beam Lines 523

Table 12.13 A gnuplot script to obtain Fig. 12.33: projected coordinates of paraxial rays along the
optical system

The IMAGE and IMAGEZ keywords allow confirming the location of the image,
which appears to be on-axis (Y .= Z .= 0), and, for respectively the horizontal and
vertical image, .∆X = −47 µm and .∆X = −13 µm upstream of the design position,
a marginal difference compared to the expected 25 cm distance from the second
quadrupole. Additional information regarding the positioning of the horizontal and
vertical images can be found under respectively IMAGE and IMAGEZ in zgoubi.res
execution listing, for instance:

Keyword, label(s) : IMAGE

RECHERCHE DU POINT DE FOCALISATION HORIZONTAL DE 10000 TRAJECTOIRES (SUR 10000)
POINT DE FOCALISATION HORIZONTAL SUR L ORBITE MOYENNE X = -4.669011E-03 CM Y = 2.017584E-05 CM
LARGEUR IMAGE, A MI-HAUTEUR = 3.363129E-04 CM, TOTALE = 1.606970E-03 CM

Keyword, label(s) : IMAGEZ

RECHERCHE DU POINT DE FOCALISATION VERTICAL DE 10000 TRAJECTOIRES (SUR 10000)
POINT DE FOCALISATION VERTICAL SUR L ORBITE MOYENNE X = -1.316689E-03 CM Z = -1.060501E-04 CM
HAUTEUR IMAGE, A MI-HAUTEUR = 3.103773E-04 CM, TOTALE = 1.311241E-03 CM

Figure 12.35 shows the horizontal and vertical phase spaces at the image plane.
They feature a wide Y-extent, due to the beam momentum spread.
(c) Using an electrostatic doublet.
The exercise can use the simulation input data file of Table 12.12, with proper values
for the E component of the field in the . E, B quadrupoles, whereas B is set to zero.
The same procedures can be applied to reproduce (a) and (b) exercises and their
outcomes.
524 12 Beam Lines

Table 12.14 Simulation input data file: case of regular magnetic quadrupole doublet, ray-trace
from object to image, .5 × 103 H+ ions. Coordinates at the image are logged in zgoubi.fai. The
INCLUDEd file Q12_hardEdge.inc is given in Table 12.12

Table 12.15 A gnuplot script to obtain Fig. 12.34: initial horizontal and vertical Gaussian distribu-
tions, and the uniform momentum distribution. The data are read from zgoubi.HISTO.out, produced
by the HISTO commands (Table 12.14)
12.4 Solutions of Exercises of This Chapter: Beam Lines 525

Fig. 12.34 From top to

bottom: Gaussian
distributions of the initial
horizontal and vertical
angles; uniform momentum
distribution as per
MCOBJET[KOBJ .= 1]
definition in Table 12.14

12.6 E, B Lens, Hard-Edge Model

(a) E and B values. Transport matrix.
The beam optics model with both E and B fields set is given in Table 12.16, and
follows the geometry given in Fig. 12.26.

Finding initial (.φ F , B F ), (.φ D , B D ) electric potential and magnetic field values
for respectively QF and QD quadrupoles is a matter of solving 2 equations with
2 unknowns, for each one of the two quadrupoles, whereas exercise 12.5 tells the
526 12 Beam Lines

Fig. 12.35 Horizontal (left) and vertical (right) phase spaces at the image plane. V-shaped, wide
extent portraits: case of regular magnetic quadrupole doublet. S-shaped portraits: case of the . E, B
doublet, the second order chromatic aberration has been canceled, only remains a third order geomet-
ric aberration typical of quadrupoles. Graphs obtained using zpop: menu 7; 1/5 to open zgoubi.fai;
2/[2,3] to select .Y, Y ' coordinates, 2[4/5] to select . Z , Z ' coordinates; 7 to plot

required combined E, B focusing strength to get proper positioning of the image plane
(25 cm downstream of QD). In fine, one needs to essentially double B compared to
the .φ = 0 case (Exercise 12.5), by virtue of Eq. 12.12.
Following this preliminary setting, a FIT then takes care of a fine adjustment of
(.φ F , B F ) and (.φ D , B D ), With focusing strengths calculated according to the hypothe-
ses of Table 12.16, namely, protons, . Bρ = 20.435 T m and thus .W = 20 keV, radius
at pole tip taken to be .a = 0.1 m, then the FIT procedure yields potentials .φ F,D and
potential and field . B F,D such that
⎧
{ ⎪ φ
⎨ K F,e = F = −46.367 m −2
φ F = −927.33 V aW
. ⇒ B ⇒ K F,m = −2K F,e
B F = 0.18950 T ⎪
⎩ K F,m = F = 92.732 m −2
a Bρ
⎧
{ ⎪ φ
⎨ K D,e = D = 68.9017 m −2
φ D = 1378 V aW
. ⇒ B ⇒ K D,m = −2K D,e
B D = −0.28159 T ⎪
⎩ K D,m = D = −137.80 m −2
a Bρ

Out of zgoubi.res execution listing, the resulting first order transport matrix .[Ri j ]
at the image is the following:

-0.142191 7.125658E-06 -1.084862E-16 -4.836514E-16 0.00000 0.00000

-1.63377 -7.03252 -1.678508E-16 -5.217183E-16 0.00000 0.00000
4.992819E-18 -4.132432E-16 -2.986418E-02 5.779779E-06 0.00000 0.00000
1.706437E-16 -2.899205E-16 -5.94705 -33.4828 0.00000 0.00000
0.00000 0.00000 0.00000 0.00000 1.00000 0.00000
0.00000 0.00000 0.00000 0.00000 0.00000 1.00000

It can be verified that . R12 = 0 and . R34 = 0, conditions for an image located 25 cm
downstream of the quadrupole doublet.
12.4 Solutions of Exercises of This Chapter: Beam Lines 527

Table 12.16 Simulation input data file: final . E, B doublet achromat, in the hard-edge quadrupole
field model. The FIT procedure adjusts the E and B field values in both quadrupoles in order to
ensure exact 25 cm image distance and zero second order chromatic coefficients

The second order matrix out of zgoubi.res execution listing shows that all chro-
matic coefficients are now zero,
1 16 1.678E-07 1 26 7.882E-07 1 36 1.517E-14 1 46 -1.116E-12 1 56 0.00 1 66 0.00
2 16 2.293E-07 2 26 7.515E-07 2 36 3.840E-14 2 46 -3.400E-12 2 56 0.00 2 66 0.00
3 16 1.631E-13 3 26 -1.349E-13 3 36 1.079E-06 3 46 6.189E-06 3 56 0.00 3 66 0.00
4 16 4.462E-13 4 26 -8.443E-13 4 36 4.015E-06 4 46 2.299E-05 4 56 0.00 4 66 0.00

(b) Point-spread function.

Now 5,000 particles are tracked through the achromat, with the goal of highlight-
ing the effect of canceled second order chromatic aberrations in the presence of
momentum spread in the beam.
The input data file is similar to that of Exercise 12.5, yet with updated .φ and . B
components (Table 12.17).
Figure 12.35 shows the horizontal and vertical phase spaces at the image plane,
where they have been superposed to those resulting from the magnetic quadrupole
doublet case for comparison. The former feature a much narrower Y-extent as a
consequence of the cancellation of the second order chromatic aberrations by the
combined . E, B field. The IMAGE and IMAGEZ keywords detail the smaller image
sizes, respectively horizontal and vertical:
528 12 Beam Lines

Table 12.17 Simulation input data file: case of an achromat quadrupole doublet, ray-tracing from
object to image, .5 × 103 H+ ions. Coordinates at the image are logged in zgoubi.fai. Details of the
potential and B field settings in Q1 and Q2 quadrupoles are given at the bottom of the Table

Keyword, label(s) : IMAGE

RECHERCHE DU POINT DE FOCALISATION HORIZONTAL DE 5000 TRAJECTOIRES (SUR 5000)
POINT DE FOCALISATION HORIZONTAL SUR L ORBITE MOYENNE X = -1.005774E-03 CM Y = -7.940402E-07 CM
DECALAGE DU CENTRE DE GRAVITE EN Y = 7.439999E-07 CM
LARGEUR IMAGE, A MI-HAUTEUR = 1.698952E-05 CM, TOTALE = 1.735729E-04 CM

Keyword, label(s) : IMAGEZ

RECHERCHE DU POINT DE FOCALISATION VERTICAL DE 5000 TRAJECTOIRES (SUR 5000)
POINT DE FOCALISATION VERTICAL SUR L ORBITE MOYENNE X = -1.803303E-03 CM Z = 6.816746E-06 CM
DECALAGE DU CENTRE DE GRAVITE EN Z = -6.865026E-06 CM
HAUTEUR IMAGE, A MI-HAUTEUR = 2.867521E-05 CM, TOTALE = 2.715059E-04 CM

12.7 E, B Lens, With Fringe Fields

Fringe fields are added, considering an unrealistically large, 5 cm bore for the sake
of enhanced possible adverse effects. Otherwise, the same geometry is maintained
(Fig. 12.26).

(a) E and B values; transport matrix

Finding the new values of the E and B components of the lenses, for a paraxial
image at the same location together with cancelled chromatic coefficients, requires
re-running the FIT procedure.
12.4 Solutions of Exercises of This Chapter: Beam Lines 529

The corresponding input file is given in Table 12.18. Here again the complete
FIT (location of the paraxial image plane concurrently with cancelled second order
chromatic aberrations) is performed in one go, taking starting field values from the
earlier matched hard-edge model.
The resulting first order transport matrix.[Ri j ] at the image, found under MATRIX
in zgoubi.res execution listing, is the following:

-0.143799 -4.091498E-08 -5.931279E-15 -1.426745E-14 0.00000 0.00000

-1.62606 -6.95399 -2.192285E-14 -5.254189E-14 0.00000 0.00000
-2.913262E-17 -1.290835E-16 -2.958046E-02 -4.487652E-08 0.00000 0.00000
1.282855E-16 6.591733E-16 -6.00354 -33.8052 0.00000 0.00000
0.00000 0.00000 0.00000 0.00000 1.00000 0.00000
0.00000 0.00000 0.00000 0.00000 0.00000 1.00000

It can be verified that . R12 = 0 and . R34 = 0, conditions for an image located 25 cm
downstream of the quadrupole doublet.
The second order matrix out of zgoubi.res execution listing shows that all chro-
matic coefficients are zero:

1 16 4.076E-07 1 26 -7.286E-07 1 36 -4.989E-10 1 46 -1.960E-09 1 56 0.00 1 66 0.00

2 16 1.408E-06 2 26 -3.008E-07 2 36 -1.796E-09 2 46 -6.978E-09 2 56 0.00 2 66 0.00
3 16 4.352E-14 3 26 9.639E-13 3 36 4.999E-07 3 46 -3.647E-07 3 56 0.00 3 66 0.00
4 16 1.149E-13 4 26 2.278E-12 4 36 4.268E-07 4 46 -7.152E-06 4 56 0.00 4 66 0.00

(b) Field and derivatives across EBMULT

. E Z (s) and . B Z (s) field components along, respectively, .(Y = 0, Z = 1 cm) and

.(Y = 1 cm, Z = 0) lines across the doublet are displayed in Fig. 12.36. For the
purpose of these graphs, fields are logged in zgoubi.plt using EBMULT[IL = 2].
(Table 12.18). Two particle trajectories are tracked for that, forced to straight lines
using OPTIONS[CONSTY ON] keyword:

’OBJET’
20.435
2
2 1
1. 0. 0. 0. 0. 1. ’Y’ ! Y0, T0, Z0, P0, X0 (unused) and D0 coordinates;
0. 0. 1. 0. 0. 1. ’Z’ ! D >> 1 would also force particles on straight trajectories,
1 1 ! in lieu of Options[CONSTY].
’PARTICUL’
PROTON
’OPTIONS’
1 1
CONSTY ON
’DRIFT’
559.
etc.
530 12 Beam Lines

Table 12.18 Simulation input data file: final doublet achromat, accounting for E and B fringe fields
(. X E = X S = 9 cm, λ E = λ S = 5 cm, . E 2 = E 3 = 1, for both E and B fields, both quadrupoles).
The FIT procedure adjusts the field values in the quadrupoles with the constraints of ensuring,
concurrently, 25 cm image distance and cancelled chromatic aberrations—MATRIX computation
follows once FIT is done, to allow checking the second order chromatic coefficients, expected null
12.4 Solutions of Exercises of This Chapter: Beam Lines 531

Fig. 12.36 Field

components along
.(Y = 1 cm, Z = 0) and
.(Y = 0, Z = 1 cm) straight
lines across the . E, B
quadrupole doublet,
simulated with EBMULT
and including fringe fields
532 12 Beam Lines

Note that an alternate way to force straight trajectories through the optical element
is to give the particles a high rigidity, e.g., . D ≈ 1010 under OBJET.
With the option EBMULT[IL = 7], field derivatives .d i+ j+k E/d X i dY j d Z k and
i+ j+k
.d B//d X i dY j d Z k across the magnetic quadrupole doublet are logged in
zgoubi.impdev.out during the step-by-step integration of motion. Some of these
derivatives are displayed in Fig. 12.37, produced using the gnuplot script given in
Table 12.19.

12.8 Animation: Dynamical Squeeze of Point-Spread Function

The input file in Table 12.20 performs a squeeze of the image of a point object with
momentum spread (uniform in .dp/ p ∈ [−10−3 , +10−3 ]). The squeeze is achieved
by slowly switching on the electric component in the quadrupoles, while slowly
decreasing the magnetic component, toward their optimum values, in 30 steps (as
per REBELOTE[NRBLT .= 30, IOPT .= 1]). IOPT .= 1 option allows the NPRM .= 4
parameters concerned to be varied, as specified by the four “KEYWORD{LABEL1}
parameter number, range” subsequent lines under REBELOTE.
The gnuplot script for this animation can be found at the bottom of Table 12.20.
It reads its data from zgoubi.fai. The animation cycles on a series of 31 images. As
an illustration, the first image (maximal size, E .= 0) and final image (minimal size,
E and B set to cancel chromatic aberrations) together with two intermediate steps,
are displayed in Fig. 12.38.

12.4.3 .π → μ + ν In-Flight Decay, Bunch Densities

12.9 Parent Pion Bunch

(a) Pion momentum density

The input data file to simulate transverse beam densities along the line is given
in Table 12.21. This is the parent pion bunch case of Fig. 12.9 with initial random
uniform . pπ ∈ [100, 500] MeV/c. The data list includes a series of field free drifts
aimed at generating the .s-dependence of pion momentum densities.
The non-decayed pion count at successive distances .s is obtained from
zgoubi.HISTO.out file generated by HISTO[PRINT], and represented in Fig. 12.39.
The latter is produced using the gnuplot script given in Table 12.22. The theoretical
pion count in a bin .[ pπ , pπ + δpπ [ writes (Eq. 12.17)
|
. δ N pπ |s = 1∆pπ ( pπ ) exp(−ηs/ pπ ) N0 δpπ (12.32)

and yields the theoretical curve .δ N pπ ( pπ ) / N0 δpπ , superposed to the tracking out-
comes in Fig. 12.39.
12.4 Solutions of Exercises of This Chapter: Beam Lines 533

Fig. 12.37 Field derivatives

along .(Y = 1 cm, Z = 0)
and .(Y = 0, Z = 1 cm) lines
across the . E, B quadrupole
doublet, simulated with
EBMULT and including
fringe fields. The two
derivatives .∂ BY /∂ Z and
.∂ B Z /∂Y are computed
independently in zgoubi,
they do superimpose, as
expected; same observation
for .∂ E Z /∂ X and .∂ E X /∂ Z ,
as well as for the three
derivatives .∂ 2 B Z /∂ X 2 ,
.∂ B X /∂ X d Z and
2

.∂ B X /∂ Z ∂ X . The interval
2

between markers is an
integration step
534 12 Beam Lines

Table 12.19 A gnuplot script to produced Fig. 12.37, field derivatives across the . E, B quadrupole
doublet

Note in passing, the decay distance from this .106 pion Monte Carlo simulation
with random uniform . pπ ∈ [100, 500] MeV/c comes out to be .< sπ >= 16.725 m,
so satisfying the theoretical expectation
ʃ∞ ʃ pπ2
0 s ds pπ1g̃ pπ ( p)|s dp
. < sπ >= ʃ∞ ʃ pπ2
0 ds pπ1 |s dp

(with .g̃ pπ ( p)|s dp from Eq. 12.17).

(b) Pion bunch momentum

The evolution of the average momentum of a pion bunch over 60 m flight distance
from Monte Carlo simulations is obtained using REBELOTE[NRBLT .= 11, IOPT
.= 1] to repeat the tracking with increasing .s. The input data file for that simulation

is given in Table 12.23. Results are displayed in Fig. 12.40.

12.10 Muon Bunch

(a) Muon momentum density at fixed . pπ . Reproducing Fig. 12.11
The input data file to simulate muon momentum densities is given in Table 12.24.
While REBELOTE[NRBLT, IOPT .= 1] repeats (NRBLT .= 2 additional passes, for a
total of three different passes, i.e. three different initial pion bunch momenta), HISTO
cumulates particle counts, pass after pass.

The muon count for various initial . pπ values is obtained from the zgoubi.
HISTO.out file generated by the HISTO keyword, and represented in Fig. 12.41
(produced using the gnuplot script given in Table 12.25), consistent with theoretical
expectations, Fig. 12.11.

(b) Muon momentum density, .s-dependence. Reproducing Fig. 12.12

The input data file to simulate the muon momentum densities along the line, as muons
decay, is derived from that of Exercise 12.9, Table 12.21, changing “40 D 0 P” to “40
12.4 Solutions of Exercises of This Chapter: Beam Lines 535

Table 12.20 Simulation input data file: case of achromat quadrupole doublet, ray-trace from object
to image,.5 × 103 H+ ions. Coordinates at the image are logged in zgoubi.fai. Details of the potential
and B field settings in these . E, B Q1 and Q2 quadrupoles are given at the bottom of the Table.
Note this file, with its INCLUDE expanded, and the gnuplot script for the animation, are available
in zgoubi sourceforge repository at https://sourceforge.net/p/zgoubi/code/HEAD/tree/branches/
exemples/book/zgoubiMaterial/beamLines/nanoProbeAchromat/animation/
536 12 Beam Lines

Fig. 12.38 A series of four successive views of the image of a point object, squeezed from a
maximal size with E .= 0, to an image of minimal size, free of chromatic aberration with optimized
E and B values. These snap shots are taken from the 30-series of the animation (Table 12.20)

D 0 S” under HISTO, for an histogram of S-econdary particles, instead of P-rimary,

and changing the drift lengths. The resulting simulation file is given in Table 12.26.
The muon count at various distances .s is obtained from the zgoubi.HISTO.out
file generated by HISTO[PRINT]. The result is represented in Fig. 12.42, produced
using the gnuplot script of Table 12.26, consistent with Fig. 12.12.

(c) Muon bunch momentum, average

The evolution of the average momentum of a muon bunch over 60 m flight distance
from Monte Carlo simulations is shown in Fig. 12.40.
It is obtained by repeating a s-distance tracking, with increasing .s up to 60 m,
using the pion case data file in Table 12.23 where “40 D 0 P” under HISTO (for
Parent particles) is changed to “40 D 0 S” (for Secondary particles).
12.4 Solutions of Exercises of This Chapter: Beam Lines 537

|
Table 12.21 Simulation input data file for the computation of . g̃ pπ ( pπ )|s density (Eq. 12.18).
Track .104 pions over 40 meters. Initial pion momentum distribution is random, uniform in
.[100, 500] MeV/c

|
Table 12.22 gnuplot script for . g̃ pπ ( pπ )|s pion density plot, reading data from zgoubi.HISTO.out
file
538 12 Beam Lines

Fig. 12.39 Pion momentum density at successive .s along the decay channel, from Monte Carlo
simulations (markers) superposed on the theoretical curves, Eq. 12.32 (solid lines). .dpπ = pπ −
pref , . pref = 250MeV/c

Table 12.23 Simulation input data file: for . p̄π (s) computation. Track .103 pions over increasing
length drift. Initial pion momentum distribution is random, uniform in .[100, 500] MeV/c
12.4 Solutions of Exercises of This Chapter: Beam Lines 539

Fig. 12.40 Average

momentum of pion and
muon bunches over 60 m
flight distance from
Monte Carlo simulations
(markers). For comparison,
superposed: solid lines from
Eq. 12.18

Table 12.24 Simulation input data file: for .g pμ | pπ ( pμ ) computation. Track .104 pions with zero
momentum spread, over a distance .s ≫ sπ , in three different cases of initial momentum . pπ =
100, 250 and .500 MeV/c
540 12 Beam Lines

Table 12.25 Gnuplot script for the plot of muon density,.g pμ | pπ ( pμ ), case of parent beam momen-
tum 100, 250 or 500 MeV/c, reading data from zgoubi.HISTO.out file

Fig. 12.41 Muon momentum density.g pμ | pπ ( pμ ), for. pπ = 100, 300 and 500 MeV/c, from Monte
Carlo simulations

Fig. 12.42 Muon momentum density at various .s along the decay channel from Monte Carlo
simulations. .dpμ = pμ − pref , . pref = 250 MeV/c
12.4 Solutions of Exercises of This Chapter: Beam Lines 541

|
Table 12.26 Simulation input data file for the computation of . g pμ ( pμ )|s . Decay of .104 pions is
tracked over 80 meters. Initial pion momentum distribution is random, uniform in.[100, 500] MeV/c

(d) Longitudinal muon phase space. Reproducing Fig. 12.13

This simulation is performed using the input data file given in Table 12.26, mutatis
mutandis, as follows:
– change the parent pion initial momentum interval to . pπ ∈ [200, 400] MeV/c,
– introduce and additional 20 m DRIFT right after DRIFT[LABEL1 .= s_20m],
– followed by FAISCNL or FAISTORE[FNAME .= zgoubi.fai] in order to get par-
ticle data at s .= 40 m in zgoubi.fai.
A graph of the longitudinal bunch population .gtμ ,Eμ (tμ , E μ )|s at .s = 40 m so
obtained is given in Fig. 12.43, consistent with Fig. 12.13.
542 12 Beam Lines

Fig. 12.43 Muon longitudinal phase-space .gtμ ,E μ (tμ , E μ )|s at .s = 40 m, from Monte Carlo sim-
ulation, and projected densities along the axes. The thin arc on the right of the muon bunch is the
survived pion bunch

12.4.4 Pion Funnel and Muon FODO Collect Channel

12.11 Beam Line Simulation

(a) Reference trajectory, transport matrix, phase space invariants.
The simulation input data file for the complete collect channel, funnel+FODO line,
described in Table 12.3, is given in Table 12.27. The data include OBJET[BORO .=
833.91, KOBJ.= 5] for transport coefficient computation by MATRIX. The reference
rigidity BORO .= 833.91 kG.cm corresponds to 250 MeV pions; however the optical
axis (i.e., the off-centered optical axis in the funnel, and the reference axis along the
center of the quadrupoles in the FODO channel) corresponds to.1.2 × 833.91 kG.cm,
or 300 MeV/c pions (hence D=1.2 relative momentum of the 13-particle set, under
OBJET[KOBJ .= 5], Table 12.27).

Figure 12.44 shows the reference trajectory along the funnel section, it is zero
beyond, along the FODO channel. Out of zgoubi.res execution listing, the resulting
first order transport matrix .[Ri j ] at the end of the line (s .= 31.1291475 m) is the
following:

Reference, before change of frame (particle # 1 - D-1,Y,T,Z,s,time) :

2.00000000E-01 3.58088321E-04 -1.40109121E-02 0.00000000E+00 0.00000000E+00 3.11291475E+03 5.47393263E-02

TRANSFER MATRIX ORDRE 1 (MKSA units)

0.157313 2.64910 0.00000 0.00000 0.00000 4.179761E-02
-0.264285 1.88869 0.00000 0.00000 0.00000 2.649820E-02
0.00000 0.00000 0.490508 2.23057 0.00000 0.00000
0.00000 0.00000 -0.560230 -0.508769 0.00000 0.00000
0.105474 -8.683260E-03 0.00000 0.00000 1.00000 -1.724655E-02
0.00000 0.00000 0.00000 0.00000 0.00000 1.00000
12.4 Solutions of Exercises of This Chapter: Beam Lines 543

Table 12.27 Simulation input data file for a beam line including an upstream pion funnel section,
and a downstream muon FODO collect channel. This sequence INCLUDEs 14 copies of the basic
cell of the FODO channel, taken from the file FODOCell_piCollect.inc (Table 12.28), in which
the segment [#S_muonFODOCell:#E_muonFODOCell] is defined using LABEL1s. This file is
resorted to in subsequent INCLUDEs under the name opticalSequence.inc. For possible further
data analysis or graphs, use OPTIONS[.plt, IL .= 2] and/or DRFIT[split, IL .= 2] to log parti-
cle data in zgoubi.plt, step-by-step as numerical integration proceeds. Note this file is available
in zgoubi sourceforge repository at https://sourceforge.net/p/zgoubi/code/HEAD/tree/branches/
exemples/book/zgoubiMaterial/beamLines/muonCollectChannel/
544 12 Beam Lines

Fig. 12.44 Reference 300 MeV/c (pions) trajectory along the funnel section of the muon collect
channel. A graph obtained using zpop: menu 7; 1/1 to open zgoubi.plt; 2/[6, 2] to select .Y versus
distance

Table 12.28 Simulation input data file for one cell of the 14-cell muon collect channel. This
is the FODO cell segment [#S_muonFODOCell:#E_muonFODOCell] subject to INCLUDE in
Table 12.27

(b) Optical functions,

They are computed and logged in zgoubi.OPTICS.out by the simulation data file
of Table 12.29. A graph is given in Fig. 12.45, it resorts to gnuplot_OPTICS.gnu, a
script found in zgoubi toolbox [31]. It is a good idea in this derivation of transport
coefficients, and optical functions, from particle trajectories, to keep an eye on the
determinants of the first order transport matrix.T (s ← 0) from s=0 to current.s, along
the line; they are displayed in Fig. 12.46 and appear to marginally differ from 1, as
12.4 Solutions of Exercises of This Chapter: Beam Lines 545

Table 12.29 Simulation input data file for the transport of optical functions along the funnel+FODO
channel. This sequence INCLUDEs the segment from occurrence of LABEL1 .= S_NF_Funnel
to occurrence of LABEL1 .= E_NF_FODO, as defined in Table 12.27. The SYSTEM command
produces a graph of the optical functions along the line (Fig. 12.45)

Fig. 12.45 Optical functions along the muon collect channel, for the reference rigidity
833.91 kG.cm (300 MeV/c pions). A graph obtained using gnuplot_OPTICS.gnu [31]
546 12 Beam Lines

Fig. 12.46 Evolution of the horizontal and vertical determinants of the first order transport matrix
. T (s
← 0), from s=0 to current abscissa .s along the line. They are computed by OPTICS, from the
13 trajectories generated by OBJET[KOBJ .= 5.1]

expected. Note that first order coefficients are computed by polynomial interpolation,
which contributes from that departure from a value of 1.
The values the optical functions at end of funnel and end of FODO channel are
given in Table 12.30.

(c) Phase space ellipses.

Motion invariants remain ellipses if particle motion is paraxial. However large
initial invariant values are considered here: .εY /π = ε Z /π = 2.5 mm, as well as large
momentum offsets (250 and 400 MeV/c are tracked, for a reference 300 MeV/c).
Distortion of the ellipse invariant along the line is expected.
OBJET[KOBJ .= 8] is used to generate the appropriate object, namely, 300 parti-
cles on.εY = 2.5 π mm invariant. In addition to a first pass at 250 MeV/c, REBELOTE
is used to repeat the tracking twice, for D .= 1.2 and 1.6 (300 and 400 MeV/c). The
simulation input data file is given in Table 12.31, resulting invariants at the end of
the muon collect channel are displayed in Fig. 12.47.

12.12 Bunch Densities

The simulation input data file for the transport of a .104 pion bunch and its daugh-
ter decay muons, along the funnel+FODO channel, is taken from Table 12.27 (using
INCLUDE), it is just required to substitute the Monte Carlo object definition MCOB-
JET[KOBJ=3] to the former OBJET[KOBJ .= 5], whereas MATRIX at the end of
the file may be removed. The resulting simulation data file is given in Table 12.32.
12.4 Solutions of Exercises of This Chapter: Beam Lines 547

Table 12.30 Optical functions at start of the line, end of funnel, and end of FODO channel, out of
zgoubi.res execution listing

(a) Sample pion and muon trajectories.

Sample pion and muon trajectories along the line are displayed in Fig. 12.48.

(b) Bunch densities.

Tracking is performed using the simulation input data file of Table 12.32. The number
of pions under MCOBJET is changed to .104 , transverse emittances are Gaussian,
cut-off at one-.σ , momentum density is uniform, in .[200, 400] MeV/c. In this file,
following the funnel,
INCLUDE[5 * opticalSequence.inc[S_NF_FODO:E_NF_FODO]]
defines a 108 m FODO channel.
The evolution of the average momentum of.π ,.μ and.π + μ beams so obtained are
displayed in Fig. 12.49; the theoretical curves, (after Eq. 12.13 for the s-dependence
of the pion density, Eq. 12.21 for the muon density) are superimposed for comparison.
The higher momentum on average from the raytracing, is due to the loss of low energy
muons in the walls defined by CHAMBR.
548 12 Beam Lines

Table 12.31 Simulation input data file for the transport, along the funnel+FODO channel, of a set of
300 particles on horizontal or vertical invariants defined (under OBJET) by.εY /π = ε Z /π = 2.5 mm
and (Tables 12.4, 12.30) .αY = α Z = 0, .βY = β Z = 4 m. This sequence INCLUDEs the segment
from occurrence of LABEL1 .= S_NF_Funnel to occurrence of LABEL1 .= E_NF_FODO, as
defined in Table 12.27

Fig. 12.47 A hundred particles on their distorted invariants at the end of the FODO channel, at
250, 300 and 400 MeV/c. A graph obtained using zpop: menu 7; 1/5 to open zgoubi.fai; 2/[2, 3] to
select .(Y, T ) phase space or 2/[4, 5] to select .(Z , P) phase space

Transverse pion bunch densities at start, 20 and 50 m, together with transverse

muon bunch densities at 100 m, are displayed in Fig. 12.50. The theoretical densities
are superimposed, for comparison.

(c) Phase space collimation.

Phase space ellipse parameters (centering and shape) which correspond to maxi-
mized acceptance at the end of the collect channel can be determined using COL-
LIMA[IFORM .= 14–16]. In the case of time-energy phase space and an ellipse
surface .εl = 0.5 π eV s (cut-off at .1σ ), COLLIMA takes the form
12.4 Solutions of Exercises of This Chapter: Beam Lines 549

Table 12.32 Simulation input data file for the transport of a bunch of 100 pions, and decay muons
as it proceeds, along the funnel followed by a 108 m FODO channel. Parent and daughter particles
are specified under PARTICUL and MCDESINT, respectively. CHAMBR defines the vacuum pipe
opening. The SYSTEM command at the bottom of the file produces a graph of Y and Z coordinates
of the particles along the line, using the gnuplot script gnuplot_Zplt_sYZ.gnu, found in zgoubi
toolbox [33]

Fig. 12.48 Horizontal and vertical trajectories of .π and .μ particles, along the collect channel. A
pion track stops when the particle either decays, or hits the vacuum pipe, whose transverse aperture
is defined by CHAMBR. A muon track stops when the particle hits the pipe
550 12 Beam Lines

Fig. 12.49 Average beam momentum of .π , .μ and .π + μ beams (relative to 250 MeV/c). Dots
represent the survived pions, at the exit of optical elements. Smooth curves are from theory, assuming
loss-free propagation. Histograms (broken lines) are from raytracing, obtained using zpop, which
reads particle data at exit of optical elements, from zgoubi.fai

Fig. 12.50 Momentum

histograms of pions at
entrance of the funnel and at
20 and 50 m, and of
(triangular shape
distribution) muon beam at
100 m. Smooth curves are
from theory. Histograms
(broken lines) are from
raytracing, obtained using
zpop: menu 7; 1/5 to open
zgoubi.fai; 2/[1,28] for an
histogram of . p/ pr e f ; 7 to
plot

’COLLIMA’
1
16 0. 0. .5 1 0. 0.

It is placed at the end of the optical sequence of Table 12.32. Figure 12.51 shows four
different ellipses so determined, with surface .εl = 0.1, 0.5, 1, and .2 π eV s.

(d) Maximization of the transmission.

It can use FIT[IC .= 5, I .= 4–6], which computes the rms ellipse that matches the
outgoing beam, and the transmission through it, while the ellipse surface is left free
(its value is an outcome of the rms match). Maximization of the transmission may
use also FIT[IC .= 5, I .= 1–3]; in that case the FIT still delivers the ellipse parameters
.αY,Z ,s and .βY,Z ,s , but the surface of the ellipse is part of the constraints.
12.4 Solutions of Exercises of This Chapter: Beam Lines 551

Fig. 12.51 Optimum ellipse

shape and positioning in
time-energy phase space, for
maximized acceptance, at
the end of the muon collect
channel, for the four cases of
emittances .εl = 0.1, 0.5, 1,
and .2 π eV s

12.4.5 Low Energy Spin Rotator

12.13 Spin Rotation, Spin Matrix

(a) E and B settings.
A 0.5 m long WIENFILT is set with magnetic and electric E and B field values proper
to ensure straight electron trajectory and.θsp = 30◦ spin Y-rotation. E and B are deter-
mined from Eq. 12.27 (straight trajectory) and from Eq. 12.30 yielding the following
contributions
( )
BZ L 1 BZ L
.θsp = (1 + aγ ) −γ + a β2 = 30◦ (12.33)
Bρ γ +1 Bρ
50.588o from BZ torque 20.588o from EY torque

given the field values

[ ( ) ]
Bρ θsp / 1
. BZ = 1 + aγ − γ + a β 2 = 0.00407378 T
L γ +1
E Y = v B Z = 982939 V/m (12.34)

The simulation input file is given in Table 12.33. It includes a FIT procedure,
aimed in this first exercise, where a hard-edge field model is used, at confirming the
theoretical E and B values.
552 12 Beam Lines

Table 12.33 Simulation input data file WFSegment_FIT.inc: .350 keV electron straight trajectory,
together with .30◦ spin rotation, across one 0.5 m segment of a 3-segment Wien filter. The initial
E and B values may be taken from Eq. 12.34, they are about doubled here for the sole purpose of
highlighting the effect of the FIT procedure. This file includes spin matrix computation, following
FIT, so only occurring when the FIT is completed. The file defines the optical sequence segment
#S_WFSegment to #E_WFSegment for INCLUDEs in subsequent exercises. Note the .C0 − C5
Enge coefficient values in the Enge fringe field model [30, WIENFILT], set to values used in
subsequent exercises. The field model is hard-edged, here, due to the fall-off parameter values
. X E = λ E E = λ B E = 0 and . X S = λ E S = λ B S = 0
12.4 Solutions of Exercises of This Chapter: Beam Lines 553

The outcome of the FIT procedure, in zgoubi.res execution listing, is as follows:

**************************************************************************************************
STATUS OF VARIABLES (Iteration # 0 / 999 max.)
LMNT VAR PARAM MINIMUM INITIAL FINAL MAXIMUM STEP NAME LBL1
6 1 11 -3.420E+06 -1.900E+06 -982938.94 -3.800E+05 0.163 WIENFILT -
6 2 12 1.600E-03 8.000E-03 4.07378127E-03 1.440E-02 6.744E-10 WIENFILT -
STATUS OF CONSTRAINTS (Target penalty = 1.0000E-15)
TYPE I J LMNT# DESIRED WEIGHT REACHED KI2 NAME LBL1
3 1 2 7 0.000000E+00 1.000E+00 2.891169E-10 Infinity MARKER #E_WFSegment
3 1 3 7 0.000000E+00 1.000E+00 1.129879E-08 Infinity MARKER #E_WFSegment
10 1 1 7 5.235988E-01 1.000E+00 5.235988E-01 Infinity MARKER #E_WFSegment
10 1 4 7 1.000000E+00 1.000E+00 1.000000E+00 NaN MARKER #E_WFSegment
10 2 4 7 1.000000E+00 1.000E+00 1.000000E+00 NaN MARKER #E_WFSegment
10 3 4 7 1.000000E+00 1.000E+00 1.000000E+00 NaN MARKER #E_WFSegment
Fit reached penalty value 4.8086E-16
**************************************************************************************************

The resulting “STATUS OF VARIABLES: FINAL” field values

. E = −982938 V/m and . B = 0.00407378 T

are in accord to high accuracy with the theoretical ones (Eq. 12.34) as expected in
this hard-dege model. A graph of the .30◦ Z-rotation of the spin in the . B⊥ plane along
a .50 cm Wien filter segment, is given in Fig. 12.52

(b) Spin matrix.

The simulation data file of Table 12.33 includes spin matrix computation. Three
electrons are needed for SPNPRT[MATRIX] to operate. All three have identical

"zgoubi.plt" u ($19==1 & $51==0 ? $14 : 1/0):(zero):($33*120):($34*120)

"zgoubi.plt" u ($19==1 & $51==1 ? $14 : 1/0):(zero):($33*120):($34*120)

0 20 40 60 80 100 120 140 160

s [cm]

Fig. 12.52 Spin motion in the . B⊥ plane, from longitudinal at s .= 0 to .30◦ to the X axis at s .=
50 cm. The thin vector series (red) is before any FIT, when field values are about twice the optimum
ones. The thick vector series (blue) is at the last pass of the FIT procedure, when fields have their
matched values E .= –982939 V/m and B .= 0.00407378 T. The vertical bars materialize the three
Wien filter segments, the first one only is tracked, here
554 12 Beam Lines

initial coordinates and rigidity, and their spins (defined under SPNTRK) have to form
a direct triedra, namely, for electrons 1, 2, 3 respectively, . S0 (1) ≡ S X , . S0 (2) ≡ SY ,
. S0 (3) ≡ S Z .
So obtained spins, spin matrix (a Z-rotation, as expected) and rotation precession
angle (.30◦ , as expected) at the end of the Wien filter segment, in zgoubi.res execution
listing, are as follows:

*******************************************************************************************************************
10 Keyword, label(s) : SPNPRT MATRIX
-- 1 GROUPS OF MOMENTA FOLLOW --
----------------------------------------------------------------------------------
Momentum group #1 (D= 1.000000E+00; particles 1 to 3 ;
Spin components of the 3 particles, spin angles :
INITIAL FINAL
SX SY SZ |S| SX SY SZ |S| GAMMA |Si,Sf| (Z,Sf_yz) (Z,Sf)
(deg.) (deg.) (deg.)
(Sf_yz : projection of Sf on YZ plane)
o 1 1.000 0.000 0.000 1.000 0.866025 -0.500000 -0.000000 1.000000 1.6849 30.000 90.000 90.000
o 1 0.000 1.000 0.000 1.000 0.500000 0.866025 0.000000 1.000000 1.6849 30.000 90.000 90.000
o 1 0.000 0.000 1.000 1.000 0.000000 -0.000000 1.000000 1.000000 1.6849 0.000 00.000 0.000
Spin transfer matrix, momentum group # 1 :
0.866025 0.500000 1.933926E-17
-0.500000 0.866025 -1.508203E-17
-2.429799E-17 3.436238E-18 1.00000
Determinant = 1.0000000000
Trace = 2.7320507888; spin precession acos((trace-1)/2) = 30.0000010762 deg
Precession axis : ( 0.00, 0.00,-1.00) -> angle to (X,Y) plane, X axis, Z axis : -90.00, 90.00, 180.00 deg
Spin precession/2pi (or Qs, fractional) : 8.3333E-02
*******************************************************************************************************************

12.14 Integration Step Size

Numerical convergence of both particle and spin motion may be affected in a non-
negligible manner if the integration step size is taken too large. On the other hand
the step size should be taken as large as tolerable to save on computation time, if
repeated thousands of electrons trials are to be performed, for statistics purposes for
instance.

(a) Perturbation of exit coordinates.

The input data file to explore the effect of step size on electron and spin coordinates
at the exit of the Wien filter, leaving field values unchanged, is given in Table 12.34.
Figure 12.53 shows that, for . E Y and . B Z maintained constant (at their theoretical
values), varying the step size up to 10 cm only has a small effect on electron position
and angle at the exit of the Wien filter segment, whereas the relative error on the spin
rotation .θs remains below .10−3 if the step size remains below 8 cm.
As a consequence, electrons can be pushed through the Wien filter in a quick 20
steps.

(b) Recovering proper coordinates, step size

The input data file to explore the effect of step size on field values, to recover null
electron coordinates and .30◦ spin rotation, is given in Table 12.35.
A scan of the step size shown in Fig. 12.54 indicates that it can be several cen-
timeters with minor impact on nominal E or B field amplitudes: these would need to
12.4 Solutions of Exercises of This Chapter: Beam Lines 555

Table 12.34 Simulation input data file: integration step size scan using REBELOTE[IOPT=1].
The focus here is the effect of the step size on the value of final position and angle of the electron
on the one hand, and of the spin rotation angle on the other hand. Fields are set to their theoretical
values, . E Y = −982938.94 V/m and . B Z = 0.00407378127 T

be tweaked by less than .10−3 (relative) compared to their theoretical values, in order
to recover 30.◦ spin rotation and straight trajectory.

12.15 Add Fringe Fields

(a) Fringe field extent; field profile; step size .∆s.
The integration region extents at both ends of WIENFILT can be fixed at . X E =
X S = 20 cm. At such large distance from the effective field boundary, given the
fringe field extents considered, .λ E and .λ B of a few centimeters, the fields from the
556 12 Beam Lines

Fig. 12.53 Final electron position (square markers) and angle (empty circles), left vertical axis,
and relative spin rotation angle error, right vertical axis, as a function of step size. . E Y and . B Z are
maintained constant, at their theoretical values

exponential Enge fall-off model are down to negligible amplitude: the loss of field
integral matters, and it is negligible in these conditions. A sample field profile along
the Wien filter, obtained using WIENFILT[IL .= 2] which logs step by step particle
data in zgoubi.plt, is displayed in Fig. 12.55.
Convergence of the numerical integration can be checked using the simulation data
file of Table 12.35, with proper fringe field parameters under WIENFILT. The latter
is INCLUDEd from Table 12.33, thus that is where . X E = 20 cm, .λ E E = 5 cm, .λ BE =
5 cm has to be substituted to . X E = λ E E = λ BE = 0, and . X S = 20 cm, .λ E S = 5 cm,
.λ B S = 5 cm to . X S = λ E S = λ B S = 0. The integration step size .∆s has to be small
enough to ensure accuracy of the numerical integration of the equation of motion
in the fringe field regions, namely .∆s < min[λ E , λ B ] (a greater .∆s value can be
used in the Wien filter body where fields are uniform, see [30, Sect. 7.10]). For this
reason, the variation range under the do-loop REBELOTE[IOPT=1] is limited to
.0.1 ≤ ∆s ≤ 2.5 cm.

Results are displayed in Fig. 12.56: if the integration step size .∆s is taken a
centimeter and below, recovering null trajectory coordinates at the ends of the Wien
filter, and exact 30.◦ spin rotation, requires adjusting E and B by less than .≈ 10−4 ,
relative.
.∆s in centimeter range allows fast computation, convenient for statistics trials:
with .∆s = 1 cm, pushing 10,000 electrons through the 3-segment, 1.5 m long Wien
filter takes about 20 seconds on a 3 GHz CPU.
12.4 Solutions of Exercises of This Chapter: Beam Lines 557

Table 12.35 Simulation input data file for a step size scan of .δ E/E and .δ B/B. This file uses the
optical sequence segment [WFSegment_FIT_S:#E_WFSegment] defined in Table 12.33. REBE-
LOTE[IOPT=1] is added and changes the step size value in WIENFILT, in NRBLT=20 steps in the
range .0.01 ≤ ∆s ≤ 10 cm. A SYSTEM call to an external gnuplot script produces a graph of the
results of this scan

(b) Unequal fall-off extents.

Different values of the electrostatic condenser and magnetic dipole gaps result in
different fringe field extents .λ E and .λ B , which in turn jeopardizes the Wien filter rule
. E Y = v B Z . However it is possible in that case, using the two knobs .δ E Y and .δ B Z , to

recover simultaneously zero trajectory coordinates at both ends of the Wien filter and
exact 30.◦ spin rotation. An input data file to simulate this is given in Table 12.36: it
performs a scan of the .λ B /λ E ratio and, using FIT, finds proper re-tuned field values.
In this simulation .λ E = 5 cm is fixed, whereas .λ B is varied between 3 and 7 cm.
The result of that scan is displayed in Fig. 12.57, where the required adjustments
of . E Y and . B Z are plotted as a function of the ratio .λ B /λ E . The “penalty” value
monitors the FIT convergence, it comes out to be .< 10−15 , ∀λ B /λ E , as requested.
Unequal E-field and B-field fall-off extents,.λ E /= λ B , cause a shift of the reference
trajectories inside the Wien filter. The reference trajectory coordinates are anyway
zeroed at the exit of the device, together with maintaining .θs = 30◦ , by adjusting E
and B amplitudes as discussed above. Such sample reference trajectories, as displayed
in Fig. 12.58, come as a sub-product of the simulation file of Table 12.36, as step by
step particle data are logged in zgoubi.plt due to the option WIENFILT[IL=2].
558 12 Beam Lines

Fig. 12.54 An integration step size scan. This graph shows, as a function of step size, the required
variation of the . E Y and . B Z fields, relative to their theoretical value, to get exact 30.◦ spin rotation
and straight trajectory

Fig. 12.55 Field shape

across a 50 cm Wien filter
segment, on-axis. The same
for both . E Y and . B Z
components here, obtained
with .λ E = λ B = 5 cm. A
graph obtained using zpop:
menu 7; 1/1 to open
zgoubi.plt; 2/[6,35] to select
. E Y versus distance and 2/[6,
32] to select . B Z versus
distance; 7 to plot

During the execution, spin data are logged to zgoubi.res listing on the one
hand, and to zgoubi.SPNPRT.Out on the other hand as an effect of the command
(Table 12.36)

’SPNPRT’ PRINT MATRIX

12.4 Solutions of Exercises of This Chapter: Beam Lines 559

Fig. 12.56 An integration step size scan, in the presence of fringe fields. This graph shows, as a
function of step size, the required variation of the. E Y and. B Z fields, relative to their theoretical value,
to get exact 30.◦ spin rotation and straight trajectory. The Enge model normalization coefficients are
taken equal, .λ E = λ B = 5 cm, in this case

There is various ways to check the spin rotation angle. One consists in a linux
‘grep ’ from zgoubi.res execution listing, namely the very line under SPNPRT which
displays the spin rotation angle of the . S X spin coordinate (which, at s .= 0, identifies
with the spin vector: . S(s = 0) ≡ S X (s = 0)), like so:

grep -n -c ’ o 1 1.000000 0.000000 ’ zgoubi.res

The result is the following sequence of 38 lines (as per REBELOTE[NRBLT .=

37], Table 12.36):

ROTATION
INITIAL COORDINATES CURRENT COORDINATES ANGLE
254: o 1 1.000 0.000 0.000 1.000 0.866025 -0.5000 -0.000 1.000 1.6849 30.000 90.000 90.000 1
342: o 1 1.000 0.000 0.000 1.000 0.866025 -0.5000 -0.000 1.000 1.6849 30.000 90.000 90.000 1
424: o 1 1.000 0.000 0.000 1.000 0.866025 -0.5000 -0.000 1.000 1.6849 30.000 90.000 90.000 1
506: o 1 1.000 0.000 0.000 1.000 0.866025 -0.5000 -0.000 1.000 1.6849 30.000 90.000 90.000 1
........
2966: o 1 1.000 0.000 0.000 1.000 0.866025 -0.5000 -0.000 1.000 1.6849 30.000 90.000 90.000 1
3048: o 1 1.000 0.000 0.000 1.000 0.866025 -0.5000 -0.000 1.000 1.6849 30.000 90.000 90.000 1
3130: o 1 1.000 0.000 0.000 1.000 0.866025 -0.5000 -0.000 1.000 1.6849 30.000 90.000 90.000 1
3212: o 1 1.000 0.000 0.000 1.000 0.866025 -0.5000 -0.000 1.000 1.6849 30.000 90.000 90.000 1
3530: o 1 1.000 0.000 0.000 1.000 0.866025 -0.5000 -0.000 1.000 1.6849 30.000 90.000 90.000 1
560 12 Beam Lines

Table 12.36 Simulation input data file: a scan of .λ B /λ E fringe field extent ratio. FIT finds the . E Y
and . B Z values which recover null trajectory coordinates at both ends of the Wien filter, together
with .30◦ spin rotation. The accuracy of the solution is quantified by the penalty, requested to be
.< 10
−15 . This input data file defines (using LABEL1’s) the optical segment #S_WFSegment_FF

to #S_WFSegment_FF for use in INCLUDE commands in a subsequent exercise

12.4 Solutions of Exercises of This Chapter: Beam Lines 561

Fig. 12.57 Variation of the. E Y and. B Z fields (relative to their hard-edge model values), required to
recover (i) exact 30.◦ spin rotation, (ii) straight trajectory, when the .λ B /λ E ratio is varied (B fringe
field extent is varied, while E fringe field extent is kept constant, here). The fit “penalty” quantifies
the distance to these two constraints, steadily small here

Fig. 12.58 A scan of the on-momentum reference trajectories across the Wien filter, with varying
ratio .λ B /λ E . The former have zero coordinates at entrance by hypothesis, and at exit (together with
◦
.θs ≡ 30 ) as a result of the FIT on E and B amplitudes. A graph obtained using zpop: menu 7; 1/1
opens zgoubi.plt; 2/[8, 2] selects .Y versus X; 7 to plot
562 12 Beam Lines

The first number in these lines is the line number in zgoubi.res execution listing.
The rotation angle of the spin, from . S(s = 0) to . S(s = 50 cm) is the 13th data in
each line, .30◦ as expected from penalty.< 10−15 at each iteration of FIT (Fig. 12.57).
A second possibility is to use zgoubi.SPNPRT.Out data, they include the spin coor-
dinates, from which the rotation angle can be computed. There exists actually a
third possibility, which is to add FAISTORE[FNAME=zgoubi.fai] in the data file, in
lieu of SPNPRT just before REBELOTE: this logs particle data in zgoubi.fai, these
include spin coordinates.

12.16 Transport of a 6D Polarized Bunch

A .104 electron bunch is considered, rms emittances .βγ εY = βγ ε Z = 10π µm,
Gaussian transverse densities truncated at 4 sigma, momentum spread uniform over
−4
.[−10 , +10−4 ].

The Wien filter model includes fringe fields and a .λ E /λ B = 7/5 ratio between
E and B fall-off extents. Adjusting the fields to . E Y = −979009.12 V/m and . B Z =
0.00406037568 T (this is actually the first case of the simulation file of
Table 12.36, where these updated values are taken from) ensures zero reference trajec-
tory coordinates at exit, straight trajectory parallel to the X axis in the Wien filter body,
and 90.o spin rotation over the 1.9 m distance (which accounts the Wien filter effective
length, 1.5 m, and the entrance and exit field integration extents . X E = X S = 0.2 m)
(Table 12.37).
Sample trajectories and fields along, obtained using zpop, are displayed in
Fig. 12.59. Phase spaces and projected densities, including spin, at the downstream
end of the 3-segment Wien filter are displayed in Fig. 12.60.

Table 12.37 Simulation input data file: launching a .104 electron bunch through the Wien filter.
Fringe fields are included
12.4 Solutions of Exercises of This Chapter: Beam Lines 563

Fig. 12.59 Sample vertical

trajectories and . E Y and . B Z
fields along, across the Wien
filter. Graphs obtained using
zpop: menu 7; 1/1 opens
zgoubi.plt; 2/[6,4] selects . Z
versus distance (or 2/[6,35]
for . E Y versus distance, or
2/[6,32] selects . B Z versus
distance); 7 to plot
564 12 Beam Lines

Fig. 12.60 Horizontal phase

space (top) and spin angle
histogram (center) at the exit
of the 3-segment Wien filter,
case of a 6D bunch with rms
emittances
.βγ εY = βγ ε Z = 10π µm,
Gaussian transverse densities
truncated at 4 sigma,
momentum spread uniform
in .±10−4 . The spin angle
spreading (bottom) stems
from the horizontal
emittance: spin angle
spreading under the effect of
.10
−4 momentum spread, and

with .εY = ε Z = 0, amounts

to .≈ ±0.01 deg
References 565

References

1. N. Tsoupas, et al., Results from the commissioning of the NSRL beam line at BNL, in Pro-
ceedings of the EPAC 2004 Conference, Lucerne, Switzerland
2. S. Meigo, et al., High power target instrumentation at J-PARC for neutron and muon source,
in Proceedings of HB2016 (WEPM2X01) (Malm_o, Sweden), pp. 391–396
3. F. Méot, Uniform, variable size rectangle beam scanning. Application to hadrontherapy. Nucl.
Instrum. Methods Phys. Res. A 564, 108–114 (2006)
4. B. Blind, Generation of a rectangular beam distribution. Report MS H811, LANL, Los Alamos,
NM 87545
5. F. Méot, T. Aniel, Non-linear tuning and halo transport in beam expanders, in Proceedings of
EPAC 1996, Sitges, Spain. http://accelconf.web.cern.ch/e96/PAPERS/MOPL/MOP074L.PDF
6. C.H. Johnson, A ring lens for focusing ion beams to uniform densities. NIM 127, 163–171
(1975); P.F. Meads, IEEE Trans. Nucl. Sci. NS-30 (1983); B. Kashy, B. Sherrill, A method for
the uniform charged particle irradiation of large targets. Nucl. Instr. Meth. B 26(4), 610–613
(1987)
7. F. Méot, T. Aniel, Principles of the non-linear tuning of beam expanders. Nucl. Instrum. Meth-
ods Phys. Res. A 379, 196–205 (1996)
8. F. Méot, T. Aniel, Calculation of non-linear envelopes in beam expanders. Phys. Rev. Spec.
Top. - Accel. Beams 3, 103501 (2000)
9. S.Y. Yavor, et al., Achromatic quadrupole lenses. Nucl. Instrum. Methods 26, 13–17 (1964)
10. F. Méot, Generalization of the Zgoubi method for ray-tracing to include electric fields. NIM A
340, 594–604 (1994)
11. Study of a European Neutrino Factory Complex, CERN NUFACT Note 122 (2002). https://
cds.cern.ch/record/610249/files/ps-2002-080.pdf
12. B. Autin, F. Méot, A. Verdier, efficiency of an alternating gradient muon collection channel
CERN NUFACT Note 128 (2003). https://cds.cern.ch/collection/Neutrino20Factory20Notes
13. B. Autin, F. Méot, Time-energy densities in π → μ decay. CERN NUFACT Note 136 (2004).
https://cds.cern.ch/record/703869/files/nufact-note-136.pdf
14. C. Iselin, H. Grote, The MAD Program. http://mad8.web.cern.ch/mad8/
15. J. Grames, et al., Two Wien filter spin flipper, in TUP025, Proceedings of 2011 Particle Accel-
erator Conference, New York, NY, USA. http://accelconf.web.cern.ch/PAC2011/papers/tup025.
pdf Figure 12.20: Copyrights under license CC-BY-3.0, https://creativecommons.org/licenses/
by/3.0; No change to the material
16. F. Méot, Spin simulations in eRHIC Wien filter. Tech. Note BNL-212123-2019-TECH, EIC/68
(2019). https://www.osti.gov/servlets/purl/1566292
17. G. Leleux, et al., Synchrotron radiation perturbations in long transport lines, in Proceedings
of the PAC 1991 Accelerator Conference. https://accelconf.web.cern.ch/p91/PDF/PAC1991_
0517.PDF
18. F. Méot, J. Payet, Numerical tools for the simulation of synchrotron radiation loss and
induced dynamical effects in high energy transport lines. Internal report CEA DSM
DAPNIA/SEA-00-01 (CEA Saclay, 2000). https://sourceforge.net/p/zgoubi/code/HEAD/tree/
branches/publications/SR/TESLA_BDS/
19. The input data file for the simulation of the 1.2 km long TESLA BDS is available at https://
sourceforge.net/p/zgoubi/code/HEAD/tree/branches/exemples/TESLA_BDS/SRalongBDS
20. Méot, F., Synchrotron radiation interferences at the LEP miniwiggler, CERN SL/94-22 (AP)
(1994)
21. F. Méot, L. Ponce, N. Ponthieu, Low frequency interference between short SR sources. PRST-
AB 4, 062801 (2001)
22. F. Méot, A theory of low frequency far-field synchrotron radiation. Part. Accel. 62, 215–239
(1999)
23. L. Ponce, R. Jung, F. Méot, LHC proton beam diagnostics using synchrotron radiation. Yellow
Report CERN-2004-007
566 12 Beam Lines

24. R. Bossart et al., Observation of visible synchrotron radiation emitted by a high-energy proton
beam at the edge of a magnetic field. NIM 164(2), 375–380 (1979)
25. R. Coïsson, Angular-spectral distribution and polarization of synchrotron radiation from a
“short” magnet. Phys. Rev. A 20, 524 (Published 1 Aug 1979); R. Coïsson, On synchrotron
radiation in non-uniform magnetic fields. Opt. Commun. 22(2), 135–137 (1977)
26. F. Méot, Mesure de profil par rayonnement ondulateur des faisceaux de protons et antiprotons.
Ph.D. Thesis. Report CERN/SPS 81-21 (ABM) 30 Oct 1981
27. Lookup https://sourceforge.net/p/zgoubi/code/HEAD/tree/branches/exemples/
SRDiagnostics/SPS_undulator
28. G4 beam line enclosure in GANIL experimental areas. Private communication (Bernard Bru,
GANIL, 2002)
29. K. Brown, A first- and second-order matrix theory for the design of beam transport systems and
charged particle spectrometers. SLAC Report-75 (1982). https://cds.cern.ch/record/283218/
files/SLAC-75.pdf
30. F. Méot, Zgoubi Users’ Guide. https://www.osti.gov/biblio/1062013-zgoubi-users-guide.
Sourceforge latest version: https://sourceforge.net/p/zgoubi/code/HEAD/tree/trunk/guide/
Zgoubi.pdf
31. gnuplot_OPTICS.gnu: a gnuplot script which plots optical functions, as read from
zgoubi.OPTICS.out. https://sourceforge.net/p/zgoubi/code/HEAD/tree/trunk/toolbox/
gnuplotFiles/ gnuplot_OPTICS/gnuplot_OPTICS.gnu
32. betaFromPlt: a post-processing Fortran tool to transport betatron functions step-by-step, using
stepwise raytracing data logged in zgoubi.plt; outputs are logged in betaFromPlt.out. A gnu-
plot_betaFromPlt.gnu script, found therein as well, plots the content of betaFromPlt.out. https://
sourceforge.net/p/zgoubi/code/HEAD/tree/trunk/toolbox/betaFromPlt/
33. gnuplot_Zplt_sYZ.gnu a gnuplot script which plots particle coordinates, etc., as read from
zgoubi.plt. https://sourceforge.net/p/zgoubi/code/HEAD/tree/trunk/toolbox/gnuplotFiles/
gnuplot_Zplt/ gnuplot_Zplt_sYZ.gnu

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0
International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing,
adaptation, distribution and reproduction in any medium or format, as long as you give appropriate
credit to the original author(s) and the source, provide a link to the Creative Commons license and
indicate if changes were made.
The images or other third party material in this chapter are included in the chapter’s Creative
Commons license, unless indicated otherwise in a credit line to the material. If material is not
included in the chapter’s Creative Commons license and your intended use is not permitted by
statutory regulation or exceeds the permitted use, you will need to obtain permission directly from
the copyright holder.
Chapter 13
Spectrometer; Mass Separator

Abstract This chapter is an introduction to nuclear physics spectrometer and mass

separator optical systems. Optical specificities are discussed, leaning on concepts and
on theoretical material introduced in earlier chapters, or found in Optical Elements
and Keywords (Chap. 14). The chapter begins with a brief overview of mass sepa-
ration techniques and their use, and continues with a reminder of underlying beam
optics principles. The simulation of mass separation requires essentially three optical
elements from zgoubi library. DIPOLE[S] which allows for curved EFBs and field
index; it also allows several, extended shimming pole plates; detector planes, in case
they are located in the fringe field region of DIPOLE[S], can be introduced using
the option IDRT (all this, inherited from the ab initio spectrometer design capabil-
ities of zgoubi). QUADRUPOLE (or MULTIPOL) to simulate focusing elements,
and DRIFT to simulate field-free sections. Beam monitoring relies on FAISCEAU,
FAISTORE, IL = 2 flag, etc. Graphic treatment is performed using zpop or gnuplot,
possibly resorted to via a SYSTEM call.

Notations Used in the Text

. A atomic mass
. B; . B0 magnetic field; reference
.B; . Bx , . B y , . Bs field vector; radial, axial, longitudinal components
. Bρ = p/q magnetic rigidity
.c velocity of light
'
. Dx , . Dx dispersion, its derivative . Dx' = d Dx /ds
EFB Effective Field Boundary
.F Lorentz force, .F = dmv/dt = q(E + v × B)
. f ; . fx , . f y focal distance; radial, axial
. M; M x,y magnification; radial, axial
.n radial field index
.q particle charge
' '
[ ]
.x , y radial and axial coordinates in the moving frame . (∗)' = d(∗)/ds
.R orbit radius in dipole field

© The Author(s) 2024 567

F. Méot, Understanding the Physics of Particle Accelerators, Particle Acceleration
and Detection, https://doi.org/10.1007/978-3-031-59979-8_13
568 13 Spectrometer; Mass Separator

.Ri j first order transport coefficients

.s path variable
.v; .v x , v y , vs velocity vector; its components in the moving frame
.α trajectory deviation
.β normalized velocity
.∆x, .∆y radial, axial excursion off a reference axis
.δp; .δp/ p momentum offset; relative
.ε x , ε y beam emittances
.ε wedge angle
.ρ local curvature radius

13.1 Introduction

Magnetic spectrometers populate the experimental areas of nuclear physics facilities

since the 1960s. They used to be large instruments of nuclear physics, as electron and
proton beams used for nuclear reactions in a target were being boosted to GeV range
by synchrotrons, linacs, microtrons. They are today the technological cathedrals of
the particle physics TeV collider facility, the LHC.
Spectrometers are aimed at momentum analyzes of reaction products from a target.
Space and time separation of the reaction products is achieved in an optical assembly.
This chapter only addresses the optical part of these systems, other components as
target and detectors are out of the scope. Nuclear physics magnetic spectrometry
assemblies sometimes include an analyzer section: a magnetic structure similar to,
possibly symmetric of the spectrometer optical system proper, whose role is to match
the parent beam to the latter, so allowing to present a cleaned, low momentum spread
parent beam at the reaction target. It took some time for multipoles to appear in nuclear
physics spectrometers, nothing beyond dodecapole whatsoever, and wide aperture
bulky magnets anyway for the purpose of geometrical acceptance. Mutipoles were
complex components to design in the early times, 1950–1960s, as little was available
in the way of computer aided magnet design—several years would still be needed
until the inverse problem, from field shape to magnet, would benefit from routinely
computing tools.
High resolution mass separators are based on similar optical methods to
spectrometers—after all, adding a detector at the focal plane of a mass separator
makes it a spectrometer. Their role is to purify beams sent to experiments, which
in some cases happen to be spectrometry. High order multipoles, electrostatic or
magnetic, are of common usage in mass separators, for the compensation of optical
aberrations, with some designs resorting to up to 20-pole lenses.
Mass separators and spectrometers today populate radioactive ion beam (RIB)
factories at accelerator facilities worldwide: CERN, FRIB, GANIL, GSI, RIKEN,
TRIUMF, ... (Figs. 13.1, 13.2, 13.3, 13.4 and 13.5).
13.2 Basic Concepts and Formulæ 569

Fig. 13.1 The recoil spectrometer SECAR at MSU NSCL, for astrophysics studies [2]. An upstream
mass separator purifies the beam

Fig. 13.2 PRISMA heavy ion spectrometer at INFN/LN Legnaro [3]

13.2 Basic Concepts and Formulæ

Mass separation in transverse space uses magnetic fields to discriminate particles

by their momentum. Determination of mass and charge of target reaction prod-
ucts is performed concurrently. Separation and discrimination methods include time
of flight measurements from hodoscopes, dE/dx energy loss. Beam size control
along these optical assemblies resorts to quadrupoles, multipole correctors, filtering,
570 13 Spectrometer; Mass Separator

Fig. 13.3 Magnetic spectrometers at GANIL. Left: SPEG, in operation since 1985, shown here with
its upstream analysis section. SPEG was designed using zgoubi, by Birien and Saby Valero [4],
early developers of the code. Right: LISE [5], used for atomic physics and isotope studies

Fig. 13.4 A 3-spectrometer

vertical assembly for
cross-section measurements
of hadron knock-out and
meson production at MAMI
microtron facility [6]
including the KAOS
spectrometer [7]

collimation. Due to their selectivity, incident background in spectrometers, such as

beam scattering on collimation systems, neutrons, fission fragments, hard radiation,
the nuclear reaction parent beam proper, can in a large measure be removed from the
desired beam or detected signal.

Magnetic Spectrometers [1]

Magnetic spectrometers are dispersive optical systems, they separate particles accord-
ing to their q/A ratio. Separation results from a difference in particle trajectory deflec-
tion
13.2 Basic Concepts and Formulæ 571

Fig. 13.5 HRS mass separator for the DESIR experimental hall at GANIL [8], under commissioning
at LP2IB. The red arrow materializes the beam path [9]

ʃ
q
.α = B y (s)ds (13.1)
p L

over paths .L through the magnetic assembly.

Spectrometers are often imaging systems, comprised of guiding and focusing
fields, designed to provide the image on a focal surface of an object which is the
beam-target interaction volume. The ‘rays’ in this imaging process are the trajectories
of the particles of interest.
The dispersion . Dx = ∆x/ δp/ p caused by dipole fields determines the distance
.∆x between images in the focal plane resulting from a relative difference .δp in
momentum. A spectrometer is characterized by its resolution

. R = ∆x/Dx (13.2)

where.∆x is the minimum necessary separation which permits to discriminate images

at the focal plane.
Spectrometers are large geometrical and momentum acceptance devices, com-
prised of large gap, wide radial extent dipoles, featuring extended field fall-offs.
Dipole EFBs may be designed with large wedge angles, as well as second and higher
order curvatures (Fig. 13.6) in order to mitigate the effect of optical aberrations on
the resolution. As a consequence the magnetic field across the dipoles is essentially
non uniform. The ability to discriminate particles resides in the accuracy on the field
integral along a trajectory, rather than on an accurate local knowledge of the field.
As a matter of fact, for a given field integral, a local field inhomogeneity .δ B y over
a small extent .∆s along a trajectory, resulting in a perturbation .δ B y ∆s/Bρ of the
572 13 Spectrometer; Mass Separator

Fig. 13.6 SPES II QDD spectrometer at SATURNE 3 GeV synchrotron [10]. CONCORDE
quadrupole, A1 and A2 dipoles are large acceptance magnets. DP, and D’P for reversed field,
are escape axes for the primary beam. A central trajectory defines a reference optical axis, from
target to focal surface

bend angle, finds its compensation .−δ B y ∆s elsewhere along the path. The global
effect of local field inhomogeneities is an offset of the image at the focal plane, and
does not affect the resolution.
Charged particles traveling in the magnetic fields are subjected to the Lorentz
force
.F = qv × B (13.3)

which determines their trajectories. Solving the problem in terms of paraxial optics
in the moving frame (Fig. 1.2), with .v = (vx , v y , vs ) the velocity vector,

v /v ≪ v, v y /v ≪ v, vs ≈ v
. x

requires defining a reference optical axis, or a set of reference optical axes, in general
trajectories at reference momenta properly distributed over the momentum accep-
tance of the optical system (Fig. 13.6). Focusing and optical aberration properties are
determined with respect to the latter, following the transport coefficient technique
reminded in Sect. 14.5.2.
The paraxial coordinate transport method, using theoretical modeling of the mag-
net fields and fringe fields, holds for a resolution of the order of .10−3 . For momentum
resolution to reach a few .10−4 and better, compensation of optical aberrations to high
order is critical, this requires raytracing: numerical resolution of the Lorentz force
equation, including the use of computed or measured 3D field maps of the magnets.
13.2 Basic Concepts and Formulæ 573

The fact that the accuracy on field measurements for the goal resolution bears on the
field integral, not on the local field, relaxes on fabrication and shimming tolerances,
and on the complexity of the magnetic measurement apparatus.
Spectrometer properties
Various parameters characterize the device, including,
– its focal surface: the surface defined by sweeping the image over the momentum
acceptance of the spectrometer (Fig. 13.6);
– the optical magnification: . M = (image width)/(object width); this applies to both
planes, horizontal and vertical;
– dispersion, . Dx = ∆x / δp/ p;
– resolution (Eq. 13.2): the smallest .δp/ p which can be measured; the limit is given
by the typical criterion .(image width) < (image offset ∆x(δp)), so that the ulti-
mate resolution is
Mx
.R ≈ × (object width)
Dx

– solid angle, of the order of .10−3 − 10−2 srd; a larger solid angle increases optical
aberrations and makes their compensation more difficult;
– momentum acceptance.∆p/ p: it can be up to a few tens of a percent; this determines
the radial extent of the focal surface;
– transverse extent of the focal surface, both planes; it determines the volume of the
spectrometer detection system;
– transport coefficients (a paraxial approach, Eq. 14.41, Sect. 14.5.2); they charac-
terize various optical properties, for instance:
– . R11 , . R33 : . Mx and . M y magnifications, respectively;
– . R12 and . R34 coefficients: they characterize the focusing; point-to-point focusing
if null;
– . R21 , . R43 : inverse of the focal distance: . R21 = −1/ f x , . R43 = −1/ f y :
– . R16 , . R26 : horizontal dispersion and its derivative; if . R26 = 0, beams with different
momenta exit the optical system parallel.
ʃ
Tailoring . B y (s) dl
ʃ
The field integral . B y (s) dl determines the resolution of a spectrometer. Whereas
reaching field uniformity beyond .10−3 ∼ a few 10−4 in a large dipole is a delicate
process, accuracy on the field integral instead can more easily reach.10−4 and beyond.
Local field fluctuations along the path do not change the field integral, nor the focusing
properties. ʃ
Diverse methods may be used to tailor . B y (s) dl, and by this means adjust focus-
ing properties and compensate optical aberrations, as follows.
At order zero:
– shimming at dipole ends, or, at a greater scale,
– introducing
ʃ pole pieces [4, Fig. 6]; they reduce the gap height, so changing
. B y (s) dl at percent level (Fig. 13.7).
574 13 Spectrometer; Mass Separator

Fig. 13.7 Pole pieces reduce

the gap height, so modifying
. B y as a function of . x

Fig. 13.8 A wedge angle

introduces a quadrupole
effect, . B y ∆l =
B y x tan ε ∝ x

At order 1:

– if the spectrometer includes (ʃquadrupoles:

) for a trajectory
ʃ at about constant excur-
sion .x in a quadrupole, .∆ B y (s) dl ≈ ∆B y dl ≈ Gx L;
– shaping the gap in dipole entrance and/or exit regions (an hyperbolic gap) for a
local field index .n (as in Fig. 9.11, Sect.
( 9.2.1).)This introduces a quadrupole effect
at magnet ends of the form . B y = B0 1 − n Rx ;
(ʃ )
– a wedge angle .ε at dipole entrance and/or exit EFB (Fig. 13.8): .∆ B y (s) dl =
B0 x tan ε (more in Sect. 14.4.1).

At order 2 and beyond, compensation of aberrations:

– shaping the gap to introduce

( sextupole and higher ) order terms in the field correc-
x2
tion, yielding . B y = B0 1 − n Rx + n(n−1)
2 R2
+ · · · ;
– as mentioned above, pole pieces modify . B y by modifying the gap (Fig. 13.7); .x k -
dependence (k = 1, 2, …) of the effect also results from shaping their boundaries;
– curving entrance and/or exit
(ʃ dipole EFBs,
) Fig. 13.9; this allows for sextupole effect
.∆ (s) = + 2
(parabolic
(ʃ curvature):
) B y dl bx cx ; octupole effect (cubic curva-
ture): .∆ B y (s) dl = bx + d x 3 ; and beyond, cf. FRIB’s SECAR spectrometer
magnets, they feature 4th degree polynomial EFB profiling (Fig. 13.10).

Trajectory reconstruction
Diverse approaches are possible to determine the angle at which a particle with given
momentum has left the target. In any case the transport coefficient method does not
allow large enough momentum and angle acceptance. The method used at SPES II
for instance, was the following [10, 11].
13.2 Basic Concepts and Formulæ 575

Fig. 13.9 Curved EFBs. Left end of the dipole: wedge and parabolic curvature,. B y ∆l = Ax + C x 2 .
Right end: wedge and cubic curvature, . B y ∆l = Dx + E x 3

Fig. 13.10 A dipole magnet whose entrance and exit EFBs have been designed using a 4th degree
polynomial modeling [2]

SPES II measured field maps were used. For a trajectory defined by its target
coordinates . p, yt , θt , φt , assuming .xt ≈ 0 (a “narrow” target), coordinates at the
focal,.x f , y f , θ f , φ f can be computed. A polynomial development in the final coor-
dinates recovers the former, to some accuracy, from the detector measurement data.
The coefficients of that interpolation polynomial were determined from trajectory
computation, using some fitting method. On the other hand, particles detected at the
focal have a large dispersion in time of flight and momentum, a consequence of the
large momentum and horizontal acceptance of the spectrometer. With the momentum
. p known from the previous parameterization (. p is given by the intersection of the

trajectory and the focal surface), and with the trajectory length dependence. L(x f , θ f )
determined in a similar approach, the time of flight along the trajectory is derived,
from time measurement using hodoscopes, with 1 ns resolution. Determination of
576 13 Spectrometer; Mass Separator

the vertical coordinate . yt from the reconstruction allowed in addition background

rejection to some extent.
Mass Separators
Mass separators purify the beam of interest from polluting products. This requires
separation of the constituents with a high mass resolution, of 1/20,000 and beyond.
The mass resolution can be defined as [8]

Dm
. R= (13.4)
2σ0 Mx + σ A

where
δx 1 δx 1
. Dm = = = Dx (13.5)
δm/m 2 δp/ p 2

is the mass dispersion, .σ0 the object width, . Mx the horizontal magnification, and .σ A
the increase of the image width due to aberrations. For this function, mass separators
comprise electrostatic elements, such as lenses for mass independent optics, or Wien
filters. Electrostatic elements are introduced in Chap. 2, the Wien filter is introduced
in Sect. 12.2.4. Charged particles traveling along these systems are subjected to the
force
.F = q(E + v × B) (13.6)

which determines their trajectories.

Much of the aforementioned optical constraints, properties, and design methods
regarding spectrometer optics hold concerning mass separators. After all, the goal is
similar in both causes: separating particles based on their mass, energy or charge.
The design in a first approach resorts to paraxial transport techniques in a sim-
ilar way to spectrometers, with limited achievement regarding mass resolution. In
order for a design to achieve high mass resolution, accurate knowledge of optical
aberrations, with both kinematic and field origins, is needed. All field defects matter:
inhomogeneities, fringe fields, overlapping fields, ..., and their knowledge is critical.
This requires resorting to computed 3D magnetic and electrostatic field maps, and
to numerical resolution of the Lorentz force Eq. 13.6.

13.3 Exercises

Note: some of the input data files for these simulations are available in zgoubi source-
forge repository at https://sourceforge.net/p/zgoubi/code/HEAD/tree/branches/
exemples/book/zgoubiMaterial/spectrometers/.
13.3 Exercises 577

Fig. 13.11 SPES II QDD spectrometer [10]. Optical axes at three different reference momenta are
represented, they run from the target on the left (‘CIBLE’), along the CONCORDE quadrupole axis,
down to the focal surface on the right. A1 and A2 are respectively 45 tons and 60 tons, including
coils

13.1 A Nuclear Physics Spectrometer, SPES II

Solution 13.1
SPES II QDD spectrometer (Figs. 13.6 and 13.11) used to be operated at SATURNE 2
(see experimental areas in Fig. 9.3) and CERN PS. Its double-D structure is aimed
at allowing the correction of optical aberrations using EFB curvatures. SPES II
was well adapted to three-body reactions, as two particles emitted in the forward
direction with close to zero degree incidence, with comparable momenta, can be
detected simultaneously at the focal surface. Three planes of wire chambers allowed
localizing particles exiting the spectrometer.
In this exercise, SPES II is installed in zgoubi. Subsequent numerical simulations
reproduce its optical properties, resolution, etc. Analytical field modeling is used here
to simulate SPES II quadrupole and dipoles, however similar simulations using field
maps (field measurements dating from the early 1970s) are available in zgoubi
examples sourceforge repository [12] (see also [13, PART C, Sect. 1]).
Entrance and exit EFBs of SPES II dipoles A1 and A2 (Fig. 13.11) are oriented
to contribute focusing, and shaped for compensation of aberrations at the focal sur-
face over the angular acceptance. The optics is such that trajectories at different
momenta with null incidence (i.e. traveling along the quadrupole axis) come out of
A2 essentially parallel over an extended momentum range.
(a) Install SPES II in zgoubi, using Fig. 13.11 and Table 13.1 data. For the simu-
lation of A1 and A2, use DIPOLE or DIPOLE-M, indifferently (the latter generates a
field map—this is transparent to the user—which can be printed out using DIPOLE-
578 13 Spectrometer; Mass Separator

M[IC = 1 or 2]). Set DIPOLE[IL = 2] to get magnetic field along the proton paths
logged in zgoubi.plt [13, cf. DIPOLE].
Regarding the simulation of A1 and A2 fringe fields: take a fall-off extent com-
mensurate with their 20 cm gap. Enge coefficients given in Sect. 14.3.3 (Eq. 14.12)
can be used.
Consider 670 MeV/c as the central momentum (so, the 670 MeV/c trajectory coin-
cides with the reference optical axis, “P.1” in Fig. 13.11).
Check the central trajectory length and exit angle from A2. Fine-tune as needed,
including actual dipole field (power A1 and A2 dipoles separately if necessary, in
this exercise). IMAGES for instance can be used to localize the waist. Or AUTOREF,
for an automatic move of the moving frame to the waist.
Using IL = 2 flag, produce a graph of the magnetic field across A1 and A2, in the
momentum regions 670 MeV/c and .±18%. Make sure there is no truncation of the
field at A1 or A2 boundaries.
Using OBJET[KOBJ = 1] and IMAGES, find the angle of the focal plane with
respect to the central reference axis, over . pref ± 18%. Give a graph of the footprint
of the focal surface in the horizontal plane. Repeat, using AUTOREF[IL = 2].
Check .(∂ x ' /∂ p/ p) value at exit of A2.
(b) Use the multi-dimensional grid Monte Carlo object MCOBJET[KOBJ = 2] to
create Gaussian beamlets, say 200 particles each, at three momenta, . p0 and . p0 ± δp
with .δp/ p0 ≪ 1, with some divergence .∆θ × ∆φ. Produce a graph of the three
images at the focal plane, for . po = pref = 670 MeV/c. Produce the horizontal phase
space portrait.
Using this object, find the resolution at the center of the focal surface for an ini-
tial beam solid angle .∆θ × ∆φ = ±50 mrad × ±100 mrad (take distance between
images equal to their half-width, as an image resolution criterion).
Determine the maximum horizontal divergence .∆θ and vertical divergence value
−4
.∆φ, proper to ensure .5 × 10 resolution at the center of the focal. Repeat at .±18%
momentum offsets.
13.2 SPES III Spectrometer
Solution 13.2
SATURNE’s SPES III is an Elbeck spectrometer, Fig. 13.12. It was used for the
study of kaon and pion rare decay channels. Although a normal conducting dipole,
it used to be powered up to saturating field, 3 Tesla, for operation up to, typically,
3.4 T m beam rigidity.
Most SPES III parameters are given in Table 13.2. Question marks have been
introduced, they are to be answered as part of the exercise, based on appropriate
simulations and their outcomes. Refer to SPES II data Table 13.1 to clarify Table 13.2
questions, as needed.
Regarding the geometrical acceptance, to be determined: horizontally, take an
estimate of the radial size of the dipole, from Fig. 13.12; vertically, take a 20 cm
dipole gap. Use CHAMBR to reject particles that hit the chamber walls.
Produce isomagnetic field lines of SPES III in a laboratory frame, as in Fig. 13.12.
DIPOLE-M[IC = 2] can be used to get a field map from its . B(R, θ ) analytical
13.3 Exercises 579

Table 13.1 SPES II parameters, in complement to Fig. 13.11. The downstream edge of A1 60.5 deg
sector is parallel to the upstream edge of A2 40 deg sector, the central optical axis is normal to both;
it results that the total deviation by A1 is 2.5 + 60.5 = 63 deg. In A2 the exit EFB combined radii
. R1 , R2 and straight sections .U1 , .U2 simulate a third degree curvature

Momentum resolution .≈ 5 × 10
−4

Momentum range % .±18

Angle acceptance .∆θ (horiz.), .∆φ (vertic.) mrad, mrad .±50, ±100

Solid angle .∆θ × ∆φ msrd .20

Horizontal magnification 0.5

Nominal field, dependences
Nominal fielda T 1.7
Central momentum MeV/c 670
Focal extent MeV/c 550–790
Angle of focal surfaceb deg .≈ 37

Focal surface curvature Small

Non-nominal field, dependences
Maximum field T 1.9
Case of 1.8 T field
– Central momentum MeV/c 707
– Focal point distancec at 800 MeV/c mm 40
Quadrupole, vertically focusing
Nominal gradient T/m –6.155
Dipole A1d
Central trajectory deg 60 + 2.5
deviation
Entrance EFB wedge deg 17.5
anglee .θ
Entrance EFB curvature cm –57.24
radii . R1 = R2
Exit EFB wedge angle .θ deg –10
Exit EFB curvature radii cm –158, –126
. R1 , R2

Dipole A2 d
Central trajectory deg 40
deviation
Entrance EFB wedge deg 20
angle .θ
Exit EFB curvature radii cm –350, +800
. R1 , R2

Exit EFB straight cm –26.5, +20.5

sections .U1 , .U2
a The
. two dipoles were powered in series. They are individually in this exercise
b With
. respect to central momentum axis
c
. To center of focal surface
d
. Zgoubi notations and sign conventions are used here [13, Fig. 9]
e
. Accounts for the 2.5 deg tilt of A1 wrt. incoming reference axis
580 13 Spectrometer; Mass Separator

Fig. 13.12 A schematic of SPES III Elbeck spectrometer, and a few groups of momenta converging
on the focal surface

Table 13.2 SPES III parameters, in complement to Fig. 13.12

Momentum resolution ?
Momentum range at maximum field ?
Angle acceptance .∆θ (horiz.), .∆φ (vertic.) mrad, mrad ?, ?
Horizontal magnification ?
Upstream drift, from target to dipole cm 77.3627
Dipole data, at 3 T
Central momentum MeV/c ?
Focal extent MeV/c ?
Angle of focal surfacea deg ?
Focal surface curvature ?
Dipole data b
Central trajectory radius cm 200
RM
Central trajectory deg 80
deviation
Entrance EFB wedge deg 0
angle .θ
Entrance EFB straight cm 0
segments .U1 , U2
Entrance EFB curvature cm –65
radii . R1 = R2
Exit EFB wedge angle .θ deg +69
Exit EFB straight cm 0, .∞
segments .U1 , U2
Exit EFB curvature radii cm 85, .∞
. R1 , R2
a With
. respect to central momentum axis
b Zgoubi
. notations and sign conventions used here [13, Fig. 9]
13.3 Exercises 581

Fig. 13.13 A schematic of LP2IB HRS for DESIR experimental hall at GANIL. The beam line is
symmetric with respect to its mid-distance transverse plane, at M

model. An alternate method consists in using OPTIONS[CONSTY] and raytracing

an appropriate set of rays in the bend plane. Superimpose pion trajectories on this
iso-B schematic.

13.3 A High-Resolution Mass Separator

Solution 13.3
Bordeaux LP2IB1 high resolution mass separator (Figs. 13.5 and 13.13), a beam
delivery system intended for the DESIR experimental hall at GANIL, has been
designed for a resolving power .∆m/m = 31, 000 for a .1π μrad m beam emittance.
The beam line is comprised of two 90 degree magnetic dipoles, and electrostatic
focusing and correction lenses, in a symmetric arrangement to minimize aberra-
tions [8]. In this exercise, the HRS is installed in zgoubi. Ion transport and sepa-
ration at the focal plane is simulated. These numerical simulations are based on the
parameters given in Table 13.3.
The beam line from object to image plane has the configuration
.OBJECT − d1 − MQ1 − d2 − MQ2 − d3 − FS1 − d4 − FQ1 − d5 − D − d6 −

.M − d6 − D − d5 − FQ2 − d4 − FS2 − d3 − MQ3 − d2 − MQ4 − d1 − IMAGE

In this sequence, d1-6 are drifts, MQ denotes a matching electrostatic quadrupole, FS
a sextupole, FQ a focusing quadrupole, both electrostatic, D a 90 deg magnetic dipole
with 36 deg wedge angles. M is an electrostatic 20-pole lens aimed at the correction
of aberrations, located mid-way between the two dipoles. The line is symmetric with
respect to the latter location, for the minimization of aberrations.
(a) Install the HRS 90 deg bending magnet in zgoubi. BEND or DIPOLE can be
used, indifferently. Take fringe fields into account (coefficients for the Enge

1 Laboratoire de Physique des 2 Infinis de Bordeaux.

582 13 Spectrometer; Mass Separator

Table 13.3 HRS-DESIR parameters. The Enge coefficients for the bend have been obtained by
matching from an OPERA field map of the magnet
Projectile
Species .
132 Sn.+

Mass keV .122.95721

Kinetic energy keV 60

.β = v/c .0.9879 × 10
−3

Beam line length cm 1004

Drift lengths, d1–d6 cm 42, 10, 55, 6, 95.5, 75
Dipolesa
Bending radius cm 85
Angle deg 90
Wedge angles deg 36
Quadrupolesb , length, voltagec
MQ1, MQ4 cm, V 18.5, –680
MQ2, MQ3 cm, V 18.5, 770
FS cm, V 11, tbdd
FQ cm, V 22, –860
M cm, V 30, 0
Enge coefficients for the fringe fields
0−5 = 0.498959, 1.911289, −1.185953, 1.630554, −1.082657, 0.318111
.
a Dipoles: .C

fringe field extent .λ = 6 cm ≈ gap size

. Quadrupoles: .C 0−5 = 0.296471, 4.533219, −2.270982, 1.068627, −0.036391, 0.022261
b

fringe field extent .λ = 4 cm ≈ bore diameter

c
. Approximate value. Setting to be determined as part of the exercise
d
. Voltage to be determined as part of the exercise

model (cf. Sect. 14.3.3) are given in Table 13.3). Produce its transport matrix.
Give a graph of the magnetic field along the reference orbit.
(b) Install the d1-MQ1-d2-MQ2-½d3 doublet section. ELMULT is used to simulate
these quadrupoles. Take fringe fields into account (Enge coefficients are given in
Table 13.3). Using FIT[2], find the voltage setting for a double focus at 1.165 m
downstream of the object; give the transport matrix.
Produce the electrostatic field along.Y = Z = 1 cm lines across the quadrupoles,
and a few .132 Sn.+ trajectories over the 1.165 m distance.
(c) Assemble the upstream half of the line, from objet to middle of multipole M.
The FQ1 lens causes the beam to diverge horizontally and converge vertically.
The horizontal divergence causes the beam to occupy the entire dipole magnet
acceptance and so maximizes mass dispersion. The combined effects of the
entrance and exit wedge angles of the dipoles produce a parallel beam in the
horizontal direction.
The focusing condition at the mid-plane is point-to-parallel in Y, and, in Z, point-
to-point and parallel-to-parallel. Find the proper MQ1, MQ2 and FQ1 voltages
for that.
13.4 Solutions of Exercises of This Chapter: Spectrometer; Mass Separator 583

(d) Assemble the complete line, from object to image focal plane. The second (sym-
metric) half of the separator refocuses the beam at the image focal plane.
Constrain horizontal and vertical magnifications to .±1 (sign to be determined)
using MQ1, MQ2, FQ1 coupled (for equal voltage) with respectively MQ4,
MQ3, FQ2.
Find the new MQs and FQs settings. Produce the transport matrices, at the inter-
mediate double-focus, at the middle of the line, and at the image plane.
Check the value of the mass dispersion . Dm , compare to expectations.
(e) Raytrace a 6000 particle object comprised of three Gaussian beamlets, centered
respectively at .∆p/ p = 0 and .∆p/ p = ±5 × 10−4 . For each beamlet, take rms
.δp/ p = 0, rms horizontal width .0.5 mm, and rms divergence .2 mrad. Take zero
vertical size.
Minimize the aberrations at the image plane, by means of the central multipole,
using/combining any of its up to 20-pole components.
Use these simulations to determine the spectrometer resolution (Eq. 13.4)
depending on M settings.
(f) Build an animation of the squeeze of the image at the final focus while the M
multipole field components are slowly incremented. Use REBELOTE[IOPT =
1] to loop on that slow increment, together with FAISTORE[zgoubi.fai] to log
particle data at each increment (gnuplot can be used for the animated graph,
reading from zgoubi.fai).

13.4 Solutions of Exercises of This Chapter: Spectrometer;

Mass Separator

13.1 A Nuclear Physics Spectrometer, SPES II

(a) Zgoubi input data file for SPES II. Central momentum/optical axis.
CONCORDE quadrupole is simulated using QUADRUPO. Fringe fields are not
accounted for here, an arbitrary choice, they could be, yet without major effect
expected.
DIPOLE-M and DIPOLE are both used, for respectively A1 and A2. Input data
only differ by the former requiring some field map meshing data (RMIN, RMAX,
etc.).
A1 and A2 dipole data necessary for DIPOLE[-M] and deduced from Table 13.1
are detailed in Table 13.4, using Zgoubi Users’ Guide notation, cf. Fig. 14.13,
Chap. 14. These data include up- and down-stream negative drift compensation for
the AT extra extent beyond 60.5 deg for A1, and beyond 40 deg for A2.
In particular, with .u = 87.5◦ the angle positioning of A1 (2.5 deg from the incom-
ing optical axis, Fig. 13.11), and accounting for the respective AT and ACENT values
for A1 and A2, one has

– A1 and A2 entrance angle .TE = −AC E N T + ω+ ,

584 13 Spectrometer; Mass Separator

Table 13.4 Parameters of SPES II A1 and A2 in zgoubi formalism

A1 A2
Field kG 17 17 (.×1.0159)
+
.ω − ω
− deg 60.5 40
.ω
+ deg 31 20
.ω
− deg –29.5 –20
AT deg 107 83
ACENT deg 50 44
RM cm 131 131
. TE deg (rad) –19 (–0.3316) –24 (–0.41888)
. TS deg (rad) 27.5 (0.47997) 19 (0.3316)
Entrance drift cm –44.48117 –58.3249
compensation
Exit drift cm –68.19428 –45.1069
compensation
. RE cm 136.496 138.5483
. RS cm 147.687 138.5483

– A1 exit angle .TS = AT − AC E N T + ω− − 2.5,

– A2 exit angle .TS = AT − AC E N T + ω− ,
– A1 entrance drift compensation amounts to .−R M sin u tan(TE − u),
– A2 entrance drift compensation: .−R M tan TE ,
– A1 and A2 exit drift compensation: .−R M tan TS ,
– A1 entrance radius . R E = R M sin u/ sin(AC E N T − ω+ + u),
– A2 entrance radius . R E = R M/ cos TE ,
– A1 and A2 exit radius . R S : expected close to . R M/ cos TS , eventually fine-tuned so
that the radial coordinates (position .Y and angle .T ) of the reference orbit come
out null at dipole exit.
The input data file resulting from these settings is given in Table 13.5.
With these geometrical settings, and with a 17 kG field in A1 and A2, together
with the initial conditions defined in OBJET[KOBJ = 1], the central momentum (set
to . pref = 670 MeV/c, here) exits A2 in the vicinity of, and almost parallel to the
reference axis.
Fine-tuning
Ultimately, a fine tuning of the field in A2, by a factor .1.0159 (SCALING factor for
DIPOLE family, Table 13.5; DIPOLE-M left unchanged with a SCALING factor
of 1.) ensures that the 670 MeV/c trajectory is close to the expected central axis
downstream of A2, and parallel to the latter (A1 and A2 can be varied, possibly in
series, to do so uncomment FIT, down the data file in Table 13.5). The following
excerpt from zgoubi.res shows that this fine tuning yields, at the image location,
.Y = −1.993 cm (distance to the central axis) and. T = 0.001 mrad (angle with respect
to the central axis):
13.4 Solutions of Exercises of This Chapter: Spectrometer; Mass Separator 585

Table 13.5 Simulation input data file SPES2_IMAGES.inc: SPES II spectrometer, compute images
at the focal surface . This file also defines the segment SPES2_S to SPES2_E for INCLUDE purposes
in subsequent exercises
586 13 Spectrometer; Mass Separator

This fine-tuning yields the following coordinates in the interval between A1 and A2
(an excerpt from zgoubi.res):
– under FAISCEAU at exit of A1:

D-1 Y(cm) T(mr) Z(cm) P(mr) S(cm)

0.0000 3.649 -4.023 0.000 0.000 3.118677E+02

– under FAISCEAU at entrance of A2:

D-1 Y(cm) T(mr) Z(cm) P(mr) S(cm)

0.0000 3.3604 -4.023 0.000 0.000 3.8354334E+02

These data show a residual radial offset: .Y ≈ 3.5 cm, a residual incidence: .T ≈
−4 mrad, with respect to the theoretical reference optical axis between A1 and A2.
A small enough difference to be ignored (the objective of the exercise is not a fine
analysis of a three-body rare decay reaction!).
Finally, from IMAGES one also gets a distance of .7.139608 m from the target (at
OBJET) to the focal plane, whereas Fig. 13.11 for the 670 MeV/c reference shows a
path length of
π
600 + 400 + 730 + 1310 (60.5 + 40)
. + 1300 + 1840 = 7167.8 mm
180
The agreement between both distances is at .< 0.5% level.
Distances from target to focal image are summarized in Table 13.6, based on
further raytracing results detailed below. A greater relative difference is observed for
the lower momentum, this may be an effect of fringe fields: lower momenta travel
a relatively longer distance in fringe field regions compared to higher momenta,
Fig. 13.14.

Table 13.6 Distance from target to image (in mm), from raytracing, first row, and from Fig. 13.11
for comparison, second row
Momentum –18% 670 MeV/c +18%
Raytracing 6787 7140 7515
Figure 13.11 6228 7168 7678
13.4 Solutions of Exercises of This Chapter: Spectrometer; Mass Separator 587

Fig. 13.14 Magnetic field across A1 and A2 dipoles, along trajectories at momenta . pref =
670 MeV/c and . pref ± 18% (respectively green/solid curve, red/dotted, blue/dashed), three hori-
zontal take-off angles from target in each case: T = 0 and .±20 mrad. Lower momenta experience a
strong field inhomogeneity through A2. A graph obtained using zpop, menu 7: 1/1 to open zgoubi.plt;
2/[6, 32] for . B Z versus .s; 7 to plot

Field along trajectories

In OBJET[KOBJ = 1], Table 13.5, add a few groups of momenta at .±18%, like so:

Set IL = 2 under DIPOLE[–M]. A graph of the stepwise field data so logged in

zgoubi.plt shows reasonable . B Z (s) across the dipole doublet, Fig. 13.14. No trun-
cation of the field at A1 or A2 boundaries is observed within an horizontal beam
opening of .±20 mrad, in the momentum regions . pref = 670 MeV/c and . pref ± 18%.
Tilt angle of the focal surface, using IMAGES
Using OBJET[KOBJ = 1], IMAGES will recognize the momentum groups and will
produce the location of their respective images.
Information concerning each momentum group includes the radial distance of its
image, Y coordinate, and distance of the latter to the plane normal to the reference,
X coordinate. The horizontal angle under FAISCEAU, .T , of trajectory number 1 is
the angle of the momentum of concern, to the central optical axis. For instance:
– case D = 1.18:
588 13 Spectrometer; Mass Separator

– case D = 0.82:

In the following OBJET, the number of groups is increased to 61 in order to get

a dense set of (X, Y) image location coordinates:

Running the input data file of Table 13.5 with these 61 momentum groups produces
the plot of Fig. 13.15 (an effect of the SYSTEM call to gnuplot, bottom of the graph).
Using AUTOREF
AUTOREF[I = 2] may be used for similar results, as it causes a positioning of the
moving frame at the location of the waist, with its longitudinal axis aligned on the
velocity vector of particle 1 [13, lookup INDEX, AUTOREF]. The simulation data
file of Table 13.5 is used, changing to OBJET[KOBJ = 5] and adding AUTOREF at
the end. MATRIX is also added following the latter, to get the. R26 first order transport
coefficients. Three different momenta are considered: . pref = 670 MeV/c and . pref ±
18%, The file is run three times (REBELOTE[IOPT = 1] could be used instead, to
repeat over a larger momentum set) with reference momentum in OBJET[KOBJ =
5] consecutively taken as 1, 1.18 and 0.82 respectively. AUTOREF and MATRIX
outcomes are as follows (excerpts):
.• Momentum . pref + 18%.

.• Momentum . pref = 670 MeV/c

13.4 Solutions of Exercises of This Chapter: Spectrometer; Mass Separator 589

Fig. 13.15 Footprint of the focal surface in the .(X, Y ) bend plane (squares), and its angle to the
central momentum axis (circles, right vertical axis)—the discontinuity in the region. X ≈ 0 is due to
both. X and.Y → 0. The gnuplot file below is used to produce this graph, taking data from zgoubi.res

• Momentum . pref − 18%.

.

This tells that

– the reference trajectories at the two momenta . pref = 670 ± 18% MeV respec-
tively, are at an angle of -0.144265986 deg and 1.590332936 deg to the . pref =
670 MeV/c central trajectory. This is a small angle, and . R26 = ∂ x ' /(∂ p/ p) ≈ 0,
small at all three momenta correlatively, as expected. This confirms the expected
property of parallel exit trajectories over .±18% momentum span at the focal plane;
590 13 Spectrometer; Mass Separator

Table 13.7 Simulation input data file: SPES II spectrometer. Images at the focal surface

– ignoring the curvature and other deformation of the focal surface, its angle to
the central ray is near

YC 54.49 − 2 YC 50.9 + 2
atan
. = atan ≈ 26◦ and atan = atan ≈ 38◦
XC 110.6 − 3.6 XC 62.8 + 3.6

on respectively the low momentum and high momentum side of the central axis, about
as expected from the first method, Fig. 13.15. A momentum scan may be performed
in a similar way, using REBELOTE[IOPT = 1] to iterate through the spectrometer,
changing the . D coordinate (relative rigidity) in OBJET at each iteration.
(∂ x ' /∂ p/ p) value at exit of A2.
.

R26 = (∂ x ' /∂ p/ p) is expected to be zero, or close, over a large momentum range.

.
This is the condition for different momenta to come out about parallel. The previous
simulation provides the answer.
Out of curiosity, a different approach for a similar outcome is given in Table 13.7.
That simulation computes a series of transport matrices over a 670 MeV/c .± 18%
range, using OBJET[5.N] (N = 20, here) concurrently with MATRIX. Each one of
the N = 20 matrices is computed with reference the trajectory of the first particle in
each 13 particle set. That particle leaves OBJET with null transverse coordinates,
which is what is required here.
13.4 Solutions of Exercises of This Chapter: Spectrometer; Mass Separator 591

Fig. 13.16 Evolution of SPES II . R26 (from target to focal surface) over . p = 670 MeV/c ± 18%

The transport matrix can be found in zgoubi.res, for each one of the 5 momenta
considered, . pref = 670 MeV/c, . pref ± 9%, . pref ± 18%. (another possibility is to
have the transport coefficients . Ri j logged to zgoubi.MATRIX.out, using MATRIX
[PRINT]). The rightmost term in the second line of a matrix is the dispersion deriva-
tive . R26 = Dx' . A gnuplot script given in Table 13.7 plots . R26 ( p/ pref ), Fig. 13.16.

(b) Spectrometer resolution.

The simulation data file of Table 13.5 can be used and modified/completed as
follows:
– to OBJET[KOBJ = 1] substitute MCOBJET[KOBJ = 2], formatted like so:

’MCOBJET’ 2335.
! Reference rigidity. 2
! Distribution on a grid. 10000
! Numger of particles. 1 1 1 1 1 1
! Uniform distributions, 0. 0. 0. 0. 0. 1.
! Central values of bins. 1 1 1 1 1 5
! Number of bins in momentum. 0. 0. 0. 0. 0. .001
! Relative spacing (Brho/BORO) between momentum bins. 0. 50.e-3
0. 50.e-3 0. 0.
! Width of bins. 1. 1. 1. 1. 1. 1.
! Sorting cut-offs (unused). 9 9. 9. 9. 9.
! For p(D) (unused). 186387 548728 472874
! Seeds.
592 13 Spectrometer; Mass Separator

Fig. 13.17 HISTO outcome, in zgoubi.res execution listing. A histogram of particle momentum at
the focal surface [13, PART C, Sect.1] [12]

– move the observation location at the image, by introducing FOCALE[XL =

distance to image] prior to HISTO;
– add HISTO keywords to get histograms of the coordinates, at that location,
printed out in zgoubi.res (Fig. 13.17).
Histograms may also be obtained using zpop, or some other graphic tool, reading
final coordinates (at the focal surface) from zgoubi.fai. The latter is generated by
adding FAISTORE[FNAME = zgoubi.fai] at the bottom of the input data file.
From these simulations, with MCOBJET[KOBJ = 2] above completed consid-
ering SPES II geometrical acceptance data (from Table 13.1), and assuming some
image separation criterion such as distance between two images .≥ image width, the
spectrometer resolution at the focal surface can be deduced. The expected result of
this simulation is .∆p/ p ≈ 10−3 ∼ 5 × 10−4 over the .±18% momentum range.

For additional details, please refer to Zgoubi Users’ Guide [13, PART C, Sect. 1].

13.2 SPES III Spectrometer

Zgoubi input data file for SPES III is given in Table 13.8. DIPOLE-M is used, it first
fabricates a field map through which Zgoubi then pushes the pions. DIPOLE-M[IC
= 2] logs that field map in zgoubi.map, which zpop can read to generate various
magnetic field graphs, including isofield lines.
Zpop both computes and plots the isomagnetic field lines. It allows superimposing
trajectories read from zgoubi.plt, outcomes are displayed in Fig. 13.18.
Filling up Table 13.2 with the missing data is left to the reader—the previous
SPES II exercise can be used as a guidance.
13.4 Solutions of Exercises of This Chapter: Spectrometer; Mass Separator 593

Table 13.8 SPES III simulation input data file. In-flight pion decay can be triggered if desired,
by uncommenting PARTICUL and adding MCDESINT[M@=muon mass] command. If so, muons
from the decay will be tracked as well

Fig. 13.18 Isomagnetic

fields lines, and pion tracks
at five different momenta,
using DIPOLE-M. A graph
obtained using zpop,
menu 7: 1/1 to open
zgoubi.plt; 2/[48, 42] for
Y(X), lab frame coordinates;
7 to plot. Superimpose field
lines using menu 8, opening
zgoubi.map as produced by
DIPOLE-M[IC = 2]

For additional details, including histograms of pion and decay muon beams at the
focal plane, please refer to Zgoubi Users’ Guide [13, PART C, Sect. 3].

13.3 A High-Resolution Mass Separator

(a) The input data file for the simulation of the 90 deg bend is given in Table 13.9,
resulting from the parameters given in Table 13.3.
Transport matrix
Transport matrix for BEND, an excerpt from zgoubi.res:
594 13 Spectrometer; Mass Separator

Table 13.9 Simulation input data file: HRS 90 deg bend using BEND (top), or using DIPOLE
(bottom). The two problems are stacked, zgoubi allows that. This sequence defines the seg-
ments HRS-BEND_S:HRS-BEND_E and HRS-DIPOLE_S:HRS-DIPOLE_E for the purpose of
INCLUDEs in subsequent exercises
13.4 Solutions of Exercises of This Chapter: Spectrometer; Mass Separator 595

Transport matrix for DIPOLE, an excerpt from zgoubi.res:

Note the agreement with theoretical values of horizontal transport coefficients in

the hard edge model (Eq. 14.7 with .φ = 90 deg deviation and .α = 36 deg wedge
angle):

r 11 = cos(φ−α)
cos α
= 0.72654 r 16 = R M(1 − cos φ) = 0.85
.
r 21 = − R M cos2 α = −0.55545
1 sin(φ−2α)
r 26 = sin φ + (1 − cos φ) tan α = 1.7265

The agreement is not as good for the vertical coefficients, namely

r 33 = 1 − φ tan α = −0.14125 r 34 = R Mφ = 1.3351

. α
r 43 = − tan
RM
(2 − φ tan α) = −0.73402 r 44 = 1 − φ tan α = −0.14125

This should improve if the correction for the fringe field extent is accounted for.
Equations 14.19 and 14.20 detail how to do that, it requires calculation of the . I1 inte-
gral, which can be performed using the Enge fringe field model for the fall-off (with
coefficients .C0 − C5 provided in the simulation data file, Table 13.9). From there,
the first order correction .ψ to the wedge angle can be included. This exercise is left
to the reader. Note that zpop has an option to compute . I1 from a field fall-off profile,
namely, menu 8 (Analysis/Graphic), sub-menu 18 (FRINGE-FIELD MATCHING).
Field along trajectories.
Set IL = 2 under BEND and DIPOLE (or, an alternate possibility, use OPTIONS[.plt,
2]). A graph of the field is given in Fig. 13.19.

(b) Adjusting the quadrupole doublet

596 13 Spectrometer; Mass Separator

Fig. 13.19 Vertical magnetic field component. B Z (s) along the reference axis and along a trajectory
launched with. Z = 2 cm, across BEND and DIPOLE.. B Z (s) curves from both magnets superimpose
well, they cannot be distinguished at this scale. The overshoot at the entrance EFB is for the. Z = 2 cm
case. A graph obtained using zpop, menu 7: 1/1 to open zgoubi.plt; 2/[6, 32] for . B Z versus .s; 7 to
plot

The input data file is given in Table 13.10. A FIT finds MQ1 ad MQ2 voltages for
point-to-point imaging.
Transport matrix, field, trajectories
Transport matrix for point-to-point imaging, an excerpt from zgoubi.res:

Fields along .Y = Z = 1 cm lines are given in Fig. 13.20. Point-to-point imaging is

shown in Fig. 13.21. These graphs are obtained from data logged in zgoubi.plt by
setting ELMULT[IL = 2].

(c) Adjusting transport at middle of multipole M.

Radial point-to-parallel focusing requires. R22 = 0. Axial point-to-point and parallel-
to-parallel focusing requires . R34 = 0 and . R43 = 0.
Before matching, the transport matrix from object to middle of M multipole,
502.0177 cm downstream, is (an excerpt from zgoubi.res):
13.4 Solutions of Exercises of This Chapter: Spectrometer; Mass Separator 597

Table 13.10 Simulation input data file MQdoublet.inc: d1-MQ1-d2-MQ2-½d3 quadrupole doublet
section. This list defines the segment HRS-MQ1-2_S:HRS-MQ1-2_E for the purpose of INCLUDEs
in subsequent exercises

Fig. 13.20 Electric field

component . E Z at constant
.Y = Z = 1 cm along MQ1
(bottom curve) and MQ2
(top curve). A graph obtained
using zpop, menu 7: 1/1 to
open zgoubi.plt; 2/[8, 35] for
. E Z versus . X ; 7 to plot
598 13 Spectrometer; Mass Separator

Fig. 13.21 Four trajectories

leaving the object with
'
.Y0 = ±10 mrad (red, largest
excursion) and
'
. Z 0 = ±10 mrad (blue). A
graph obtained using zpop,
menu 7: 1/1 to open
zgoubi.plt; 2/[6, 2] (and 2/[6,
4]) for .Y (and . Z ) versus .s;
7 to plot

R22 is nearly zero, . R34 and . R43 are small, a good starting point for the matching
.
procedure. This is done running the input data file, Table 13.11. The FIT status
comes out to be (an excerpt from zgoubi.res):

It shows convergence toward the expected . R22 = 0, . R34 = 0 and . R43 = 0. This is
obtained with quadrupole voltages:

. M Q1 : −694.57599 V, M Q1 : 839.78984 V, F Q1 : −909.05344 V

The resulting transport matrix is (an excerpt from zgoubi.res):

13.4 Solutions of Exercises of This Chapter: Spectrometer; Mass Separator 599

Table 13.11 Simulation input data file objetToM.dat: from the object plane to the middle of multi-
pole M. The FIT procedure finds the proper MQ1, MQ2 and FQ1 voltages for. R22 = 0,. R34 = 0 and
. R43 = 0. The FS-FQ.inc file INCLUDEd here is comprised of (Table 13.3) ½d3-FS1-d4-FQ1-d5

(d) Adjusting transport at final focus.

A full HRS line is obtained by prolonging the upstream half, Table 13.11, by its
reverse. Now constrain . R11 , R22 to -1 and . R33 , R44 to 1, using, as variables, MQ1,
MQ2, FQ1 coupled with respectively MQ4, MQ3, FQ2. The FIT data needed for
that are:
600 13 Spectrometer; Mass Separator

With these variables and constraints, the final status of the FIT comes out to be:

as expected. It tells that this requires the following adjusted quadrupole voltages:

. M Q1, M Q4/ − 694.58, M Q2, M Q3/ + 839.79, F Q1, F Q2/ − 909.05

Following from these voltage adjustments, the transport matrix at the first focus
is still near point-to-point focusing, both planes:

Transport matrix at the middle of the multipole still features near-zero . R22 , R34 ,
R43 :
13.4 Solutions of Exercises of This Chapter: Spectrometer; Mass Separator 601

The matrix at final focus features the expected. R11 = R22 = −1 and. R33 = R44 =
+1:

From this matrix, the mass dispersion (Eq. 13.5) comes out to be

1
. Dm = Dx = −33.6 cm/%
2
fairly close to the expected . Dm ≈ 31 cm/% design value [8].
(e) A 6000 particle initial object. Minimizing second order aberrations.
To track a 6000 particle initial object, in the previous simulation for the transport
coefficients, change OBJET[KOBJ = 5] to a Monte Carlo object definition. Its format-
ting and functioning are detailed in [13, cf. MCOBJET, Sect. 6.2]. Add as well FAIS-
TORE[FNAME = zgoubi.fai FF], where FF is a LABEL1 (as in “‘MARKER’ FF”)
placed at the location of the image plane, end of the line. The resulting input data file
is given in Table 13.12. It grabs from the HRSComplete.inc file (the INCLUDE state-
ment) the segment [HRS-MQ1-2_S:FF] which is the complete HRS line sequence
(the same as used in question (d)).
Outcomes are given in Fig. 13.22. These results are actually obtained with appro-
priate settings of the mid-plane multipole M, which reduce the sextupole aberration.
Finding the necessary M component is left to the reader; FIT[IC = 3 or 4] can be used

Table 13.12 Input data file: tracking 6,000 .Sn132+ ions from object to final image plane
602 13 Spectrometer; Mass Separator

Fig. 13.22 At final focus of HRS-DESIR, for three different momenta, . p = 121.46962 ± 5 ×
10−4 MeV/c: horizontal phase space (Y’,Y), vertical phase space (Z’Z) and .(Y, Z ) transverse cross
section. These results are obtained accounting for a second degree EFB curvature (a sextupole
effect, Fig. 13.9) at both EFBs of both 90 deg bends

Fig. 13.23 Horizontal phase space for the central momentum, at final focus of HRS-DESIR, with
high order aberrations progressively minimized. The process starts from initial condition where M
is not used (curve with greater Y excursion). Minimization involves sextupole, octupole, decapole,
dodecapole and 20-pole components of M

for that. Note that similar simulations may be found in zgoubi sourceforge reposi-
tory, https://sourceforge.net/p/zgoubi/code/HEAD/tree/branches/exemples/ folder.
Minimization of high order aberrations is pushed further in Fig. 13.23: the sex-
tupole, octupole, decapole, dodecapole and 20-pole components of M are set to
minimize the image size. The initial object in this case is comprised of 300 particles,
and only has horizontal divergence (initial Y, Z and P of particles are zero).

(f) Animation of the image squeeze.

An excerpt of an animation of the image squeeze is displayed Fig. 13.23. The squeeze
relies essentially on the use of REBELOTE[IOPT = 1], which loops over varying
References 603

values of a set of non-linear field components in the M multipole. Zgoubi input files
can be found at [14]. The animation reads its data from a zgoubi.fai style file.
The files for that simulation are available in zgoubi sourceforge repository at
https://sourceforge.net/p/zgoubi/code/HEAD/tree/branches/exemples/book/
zgoubiMaterial/spectrometers/HRS-DESIR_massSeparator/animation.

References

1. J. Saudinos, Spectrométrie magnétique. Ecole d’été des Houches “Méthodes expérimentales

en physique nucléaire” (1974)
2. F. Bødker et al., Magnets and Wien filters for SECAR, in Proceedings of IPAC2017 (Copen-
hagen, Denmark TUPVA051). Figure 13.1: copyrights under license CC-BY-3.0, https://
creativecommons.org/licenses/by/3.0; no change to the material
3. Lezione 10: Magnetic Spectrometers. INFI LN-Legnaro, Italy. Figure 13.2: credit Dr. L. Corradi
and Dr. E. Fioretto, INFN-LNL
4. L. Bianchi et al., SPEG: An energy loss spectrometer for GANIL. NIM A 276(3), 509–520
(1989)
5. L. Perrot et al., Status of the CAVIAR detector at LISE-GANIL. https://accelconf.web.cern.ch/
HIAT2009/papers/g-02.pdf Figure 13.3, LISE: copyrights under license CC-BY-3.0, https://
creativecommons.org/licenses/by/3.0; no change to the material
6. R. Heine et al., Extension of the 3-spectrometer beam transport line for the KAOS spec-
trometer at MAMI and recent status of MAMI, in MOPZ037 Proceedings of IPAC2011 (San
Sebastián, Spain). Figure 13.4: copyrights under license CC-BY-3.0, https://creativecommons.
org/licenses/by/3.0; no change to the material
7. A. Tkatchenko, F. Méot, Calculs optiques pour le spectromètre à kaons de GSI. Rapport Interne
CEA/LNS/GT/88-07, Saclay (1988)
8. T. Kurtukian-Nieto et al., SPIRAL2/DESIR High Resolution Mass Separator. NIM B Volume
317, Part B, 2013, pp. 284–289. https://accelconf.web.cern.ch/ipac2021/papers/mopab264.pdf
9. J. Michaud et al., Commissioning of the DESIR high-resolution separator at CENB-G, in
12th International Particle Accelerator Conference IPAC2021 (Campinas, SP, Brazil). Fig-
ures 13.5, 13.13: copyrights under license CC-BY-3.0, https://creativecommons.org/licenses/
by/3.0; no change to the material
10. J. Thirion, P. Birien, Le Spectromètre II. Rapport Interne DPh-N ME, CEA Saclay (1975)
11. H. Catz, Le Spectromètre SPES II (CEA Saclay, circa, Rapport Interne DPh-N ME, 1975)
12. SPES II input data files for Zgoubi simulations, using the early 1970s measured field maps of
the three magnets, are available at https://sourceforge.net/p/zgoubi/code/HEAD/tree/branches/
exemples/spectrometers/spes2_spectrometer/usingFieldMaps/
13. F. Méot, Zgoubi Users’ Guide. https://www.osti.gov/biblio/1062013-zgoubi-users-guide. An
up-to-date version of the guide can be found at: https://sourceforge.net/p/zgoubi/code/HEAD/
tree/trunk/guide/Zgoubi.pdf
14. An animation of DESIR-HRS image squeeze using gnuplot, reading data from a zgoubi.fai style
file, can be found at https://sourceforge.net/p/zgoubi/code/HEAD/tree/branches/exemples/
CENBG/HRS-DESIR/animationAtFinalFocus/
604 13 Spectrometer; Mass Separator

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0
International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing,
adaptation, distribution and reproduction in any medium or format, as long as you give appropriate
credit to the original author(s) and the source, provide a link to the Creative Commons license and
indicate if changes were made.
The images or other third party material in this chapter are included in the chapter’s Creative
Commons license, unless indicated otherwise in a credit line to the material. If material is not
included in the chapter’s Creative Commons license and your intended use is not permitted by
statutory regulation or exceeds the permitted use, you will need to obtain permission directly from
the copyright holder.
Chapter 14
Optical Elements and Keywords,
Complements

Abstract This chapter is not a review of the .60+ optical elements of zgoubi’s
library. They are described in the Users’ Guide. One aim here is, regarding some of
them, to briefly recall some aspects which may not be found in the Users’ Guide and
yet addressed, or referred to, in the theoretical reminder sections and in the exercises.
This chapter is not a review of the .40+ monitoring and command keywords available
in zgoubi, either. However it reviews some of the methods used, by keywords such
as MATRIX (computation of transport coefficients from sets of rays), FAISCEAU
(which produces beam emittance parameters), and others. This chapter in addition
recalls the basics of transport and beam matrix methods, in particular it provides the
first order transport matrix of several of the optical elements used in the exercises, in
view essentially of comparisons with transport coefficients drawn from raytracing,
in simulation exercises.

14.1 Introduction

Optical elements are the basic bricks of charged particle beam lines and accelerators.
An optical element sequence is aimed at guiding the beam from one location to
another while maintaining it confined in the vicinity of a reference optical axis.
Zgoubi library offers of collection of about 100 keywords, amongst which about
60 are optical elements, the others being commands (to trigger spin tracking, trig-
ger synchrotron radiation, print out particle coordinates, compute beam parameters,
etc.). This library has built over half a century, so it allows simulating most of the
optical elements met in real life accelerator facilities. Quite often, elements avail-
able provide different ways to model a particular optical component. A bending
magnet for instance can be simulated using AIMANT, or BEND, CYCLOTRON,
DIPOLE[S][-M], FFAG, FFAG-SPI, MULTIPOL, QUADISEX, or a field map and
TOSCA, CARTEMES or POLARMES to handle it. These various keywords have
their respective subtleties, though, more on this can be found in the “Optical Elements
Versus Keywords” Section of the guide [1, pp. 12, 227], which tells “Which optical
component can be simulated. Which keyword(s) can be used for that purpose”. For a

© The Author(s) 2024 605

F. Méot, Understanding the Physics of Particle Accelerators, Particle Acceleration
and Detection, https://doi.org/10.1007/978-3-031-59979-8_14
606 14 Optical Elements and Keywords, Complements

complete inventory of optical elements, refer to the “Glossary of Keywords” found

at the beginning of PART A [1, p. 9] or PART B of the Users’ Guide [1, p. 229].
Optical elements in zgoubi are actually field models, or field modeling methods
such as reading and handling field maps. Their role is to provide the numerical
integrator with the necessary field vector(s) to push a particle, and possibly its spin,
along a trajectory. The following sections introduce the analytical field models which
the simulation exercises resort to.
Zgoubi’s coordinate nomenclature, as well as the Cartesian or cylindrical refer-
ence frames used in the optical elements and field maps, have been introduced in
Sect. 1.2 and Fig. 1.5.

14.2 Drift Space

This is the DRIFT, or ESL (for the French “ESpace Libre”) optical element, through
which a particle moves on a straight line. From the geometry and notations in
Fig. 14.1, with . L the length of the drift, coordinate transport satisfies

 X f − Xi = L

 Y f − Yi = L tan T
. (14.1)
 Z f − Z i = L tan P/ cos T

 path length d = L/(cos T cos P)

Linear approach
Coordinate transport from initial to final position in the linear approximation is
written (with .z standing indifferently for .x or . y, subscripts i for initial and f for final
coordinates) (Fig. 14.2)

Fig. 14.1 An L-long drift in

zgoubi (O;X,Y,Z) frame,
with origin at the upstream
end of the drift. A particle
flies from . A(Yi , Z i ) to
. B(Y f , Z f ), at an angle . P to
the .(X, Y ) plane. Projection
W of its straight path in
.(X, Y ) plane is at an angle . T
to the X axis
14.3 Guiding 607

Fig. 14.2 A drift section

with length . L = s f − si , and
projection of a straight
trajectory in the .(s, z) plane,
at an angle .z  (standing for .x 
or . y  ) to the .s axis

⎛ ⎞
 1 L 00 0 0
 z f = zi + L z  ⎜0 1 00 0 0 ⎟
  i
⎜ ⎟
 z = z ⎜0 0 10 L 0 ⎟
 f i
⎜ ⎟
.
 δl f − δli = βcδt = L δp or, Tdrift = ⎜0
⎜
0 00 1 0 ⎟
⎟
(14.2)
 γ2 p ⎜ L ⎟
 ⎝0 0001 2 ⎠
 δp f / p = δpi / p γ
00000 1

where .βc is the particle velocity, . p = γ mβc its momentum, .γ is the Lorentz rela-
tivistic factor.

14.3 Guiding

Beam guiding is in general assured using dipole magnets to provide a field vector
normal to a bend plane. Gradient dipoles combine guiding and focusing in a sin-
gle magnet, this is the case in cyclotrons where the field index is tailored to ensure
isochronism, in scaling FFAGs where. B ∝ r k ensures the zero-chromaticity property.
This may also be the case in strong focusing synchrotrons, for instance in the BNL
AGS [2], in the CERN PS [3]. Dipole magnets sometimes include a sextupole com-
ponent for the compensation of chromatic aberrations [4]. Non-linear optical effects
may be introduced in addition by shaping entrance and/or exit EFBs, a parabola
for instance for .x 2 field integral dependence, a cubic curve for .x 3 dependence (see
Chap. 13).
Low energy beam guiding also uses electrostatic deflectors, shaped to provide a
field normal to the trajectory arc, and possibly focusing properties. Plane condensers
may be used as well for beam steering, including beam filtering in combination with
a magnetic field, and at high energy in addition for such functions as pretzel orbit
separation, extraction septa, etc.
608 14 Optical Elements and Keywords, Complements

Guiding optical elements are dispersive systems: trajectory deflection has a first
order dependence on particle momentum.

14.3.1 Dipole Magnet, Curved

This is the DIPOLE element (an evolution of the 1972s AIMANT [1]) or variants:
DIPOLES, DIPOLE-M. Lines of constant field in the magnet body are isocentric
circle arcs. The magnet reference curve is a particular arc, at a reference radius . R0
for which the field value is . B0 . The field in the median plane can be written

2 3
r − R0 r − R0 r − R0
. B Z (r, θ) = G(r, θ) B0 1+ N + N + N  + ···
R0 R0 R0
(14.3)
. N (n) = d n N /dY n are the field index and derivatives. .G(r, θ ) describes the azimuthal
shape of the field, from a plateau value in the body to zero away from the magnet. It
can be written under the form [5]

. G(r, θ ) = G 0 F(d(r, θ )) (14.4)

where.G 0 a constant factor, and. F(d) a convenient model for the field fall-off, such as
the Enge model discussed in Sect. 14.3.3. In that model take .d(r, θ ) the distance from
particle location .(X, Y, Z ) to the magnet EFB, .λ(r ) an .r -dependent characteristic
extent of the field fall-off (e.g., representing a radial dependence of dipole gap height
.gap(r ), such that .λ(r ) ≈ gap(r )). The latter allows modeling the .r -dependence of
the flutter and its effect on vertical focusing.
Linear approach
The first order transport matrix of a sector dipole with curvature radius .ρ, deflection
α and index .n, in the hard-edge model, writes
.

⎛ ⎞
r x2 ⎡
Cx Sx 0 0 0 ρ (1 − C x ) C = cos ρα
⎜ ⎟ r
−S
⎜ C x Sx 0 0 0 ⎟1 ⎢ C  = dCds = ρ dα = r 2
1 dC
⎜ ⎟ρ Sx ⎢
⎜ 0 0 Cy Sy 0
0 ⎟ ⎢ S = r sin ρα
. Tbend =⎜
⎜
⎟ with ⎢
⎟
r
⎢ S = d S = 1 d S = C (14.5)
⎜ 0 0 C y S y 0
0 ⎟ ⎢ ρ dα
⎣ (∗) : r = ρ/√1 − n
ds
⎜1 r x2 r x3 ⎟
⎝ ρ Sx ρ (1 − C x ) 0 0 1 ρ 2 (ρα − Sx ) ⎠
x √
0 0 0 0 0 1 (∗) y : r = ρ/ n

or, explicitly,
14.3 Guiding 609
⎛ √ √ √ ⎞
cos 1 − nα √ ρ sin 1 − nα 0 0
ρ
0 1−n (1 − cos 1 − nα)
⎜ √1−n
1−n
√ √ √ ⎟
⎜− √ 1 sin 1 − nα ⎟
⎜ ρ sin 1 − nα cos 1 − nα 0 0 0
1−n ⎟
⎜ √ ρ √ ⎟
⎜ 0 0 cos nα √ sin nα 0 0 ⎟
.Tbend =⎜
⎜ √ n ⎟
⎟ (14.6)
⎜ n √ √ ⎟
⎜ 0 0 − ρ sin nα cos nα 0 0 ⎟
⎜ √ √ ⎟
⎝ √ 1 sin √1 − nα ρ
(1 − cos
√
1 − nα) 0 0 1 ρ 1−nα−sin 1−nα ⎠
1−n 1−n (1−n) 3/2
0 0 0 0 0 1

Cancel the index in the previous sector dipole, introduce a wedge angle .ε at entrance
and exit EFBs, introduce the flutter term .ψ to account for dependence of vertical
focusing on fringe field extent (see Sect. 14.4.1, Eq. 14.20). The first order transport
matrix, accounting for the entrance and exit EFB wedge focusing, then writes
⎛ cos(α−ε) ⎞
cos ε ρ sin α 0 0 0 ρ(1 − cos α)
⎜ − sin(α−2ε) cos(α−ε) 0 sin(α−ε)+sin ε ⎟
⎜ 2
ρ cos ε cos ε 0 0 cos ε ⎟
⎜ ⎟
⎜ 0 0 1 − α tan(ε − ψ) ρα 0 0 ⎟
.Tbend = ⎜ ⎟ (14.7)
⎜ − tan(ε−ψ) (2 − α tan(ε − ψ)) 1 − α tan(ε − ψ) ⎟
⎜ 0 0 ρ 0 0 ⎟
⎝ ⎠
sin α 0 0 0 1 ρ(α − sin α)
0 0 0 0 0 1

14.3.2 Dipole Magnet, Straight

This is the MULTIPOL element. Lines of constant field in the magnet body are
straight lines. An early instance of a straight dipole magnet is the AGS main dipole
(Fig. 9.2), which combines steering and focusing, and features in addition a small sex-
tupole defect component [7]. The multipole components . Bn (X, Y, Z ) [n = 1 (dipole),
2 (quadrupole), 3 (sextupole), …] in the Cartesian frame of the straight dipole derive,
by differentiation, from the scalar potential
⎛ ⎞⎛  π ⎞
∞
 G (2q) (X )(Y 2 + Z 2 )q 
n sin m Y n−m Z m
. Vn (X, Y, Z ) = (n!) ⎝
2
(−1)q ⎠⎜
⎝
2 ⎟
⎠
4q q!(n + q)! m!(n − m)!
q=0 m=0
(14.8)
where .G (2q) (X ) = d 2q G(X )/d X 2q . In the case of pure dipole field for instance

G  (X ) 2 G (4) (X ) 2
. V1 (X, Y, Z ) = G(X ) Z − (Y + Z 2 ) + (Y + Z 2 ) Z ... (14.9)
8 512
and
610 14 Optical Elements and Keywords, Complements

∂ V1 G  (X ) 2
. B X (X, Y, Z ) = − = G  (X ) Z − (Y + Z 2 ) ...
∂X 8
∂ V1 G  (X ) G (4) (X )
BY (X, Y, Z ) = − =− Y+ Y Z ..
∂Y 4 256
∂ V1 
G (X ) (4)
G (X ) 2
B Z (X, Y, Z ) = − = G(X ) − Z+ (Y + 3Z 2 ) ... (14.10)
∂Z 4 512

The longitudinal form factor .G(X ) accounts for the field fall-offs at the ends of the
magnet, it is modeled using the Enge model discussed in Sect. 14.3.3.

14.3.3 Fringe Field, Modeling, Overlapping

A fringe field model is described here, which is resorted to in several optical elements
of zgoubi’s library.
Field shape at the EFBs of magnetic or electrostatic devices can be simulated
using a hard-edge model (the field is assumed to change following a Heaviside step).
When using stepwise ray-tracing techniques however, a smooth change of the field
can accurately be accounted for. An efficient model is Enge’s field form factor [6]

1
. F(d) = (14.11)
1 + exp P(d)
2 3 4 5
d d d d d
. P(d) = C0 + C1 + C2 + C3 + C4 + C5
λ λ λ λ λ

where .d is the distance to the field boundary, and .λ ≈ gap aperture is the extent of
the fall-off. The latter is normally commensurate with gap aperture in a dipole, or
.r pole tip /(n − 1) in a multipole (.n = 2, 3, . . . for quadrupole, sextupole...).
As an illustration, Fig. 14.3 shows . F(d) as matched to the measured end fields of
BNL AGS main magnet [8, 9], using

λ = gap aperture ≈ 10 cm and

. (14.12)

. C0 = 0.45473, C1 = 2.4406, C2 = −1.5088, C3 = 0.7335, C4 = C5 = 0

These .Ci coefficient values result from an interpolation to measured field data, which
are also represented in the figure. The location of the EFB results from the following
constraint, which is part of the matching: the field integral on the down side of the
fall-off (the region from A to X .= 0 in Fig. 14.3) is equal to the complement to 1 of
the field integral on the rising side of the fall-off (X .= 0 to B region in the figure),
which writes
14.3 Guiding 611

Fig. 14.3 Longitudinal field form factor .G(X ) (normalized to one) in BNL AGS main bend, taken
along the magnet reference axis. Solid line: from Eqs. 14.11 and 14.12; square markers: measured
field data. . X = 0 is the origin in the field map frame, the vertical dashed line at . X EFB = −5.62 cm
is the location of the EFB

 X EFB  XB  B  B
. F(X ) d X = dX − F(X ) d X ⇒ X EFB = X B − F(X ) d X
XA XEFB XEFB A
(14.13)
A convenient property of this model is that changing the slope of the fall-off (i.e.,
changing .λ) will not affect the location of the EFB.
Inward fringe field extents may overlap when simulating an optical element
(Fig. 14.4). A way to ensure continuity of the resulting field form factor in such
case is to use
. F = FE + FS − 1 or F = FE ∗ FS (14.14)

where . FE (. FS ) is the entrance (exit) form factor and follows Eq. 14.11. Both expres-
sions can be extended to more than two EFBs (for instance 4, to account for the
4 faces of a dipole magnet: entrance and exit faces, inner and outer radial bound-
aries). Note that in that case of overlapping field extents, the field integral is affected,
decreasing with more pronounced overlapping, it is therefore necessary to change
the field value (. B0 in Eq. 14.4 for instance) to recover the proper integrated strength.

Overlapping Fringe Fields

Zgoubi allows a superposition technique to simulate the field in a series of neighbor-

ing magnets. The method consists in computing the mid-plane field at any location
.(r, θ ) by adding individual contributions, namely [5]
612 14 Optical Elements and Keywords, Complements

Fig. 14.4 A sketch of

overlapping entrance field
form factor . FE (d E ) (at the
entrance “EFB-E”) and exit
. FS (d S ) (at the exit
“EFB-S”). The resulting
form factor . F = FE × FS is
actually accounted for in
modeling the field

 
. B Z (r, θ ) = BZ,i (r, θ ) = BZ,0,i Fi (r, θ ) R i (r )
i=1,N i=1,N

∂ k+l B Z (r, θ )  ∂ k+l BZ,i (r, θ )

= (14.15)
∂θ ∂r
k l
i=1,N
∂θ k ∂r l

with .Fi (r, θ ) and .R i (r ) taken independently for each individual dipole in the series
(for instance as per Eqs. 10.7 and 10.15). Note that, in doing so it is not meant that
field superposition would apply in reality (if magnets are closely spaced, cross-talk
may occurs), however it appears to allow closely reproducing magnet computation
code outcomes.

Short Optical Elements

In some cases, an optical element in which fringe fields are taken into account (of
any kind: dipole, multipole, electrostatic, etc.) may be given small enough a length,
. L, that it finds itself in the configuration schemed in Fig. 14.4: the entrance and/or the
exit EFB field fall-off extends inward enough that it overlaps with the other EFB’s
fall-off. In zgoubi notations, this happens if . L < X E + X S . As a reminder [1]: in
the presence of fringe fields, . X E (resp. . X S ) is the stepwise integration extent added
upstream (resp. added downstream) of the actual extent . L of the optical element.
In such case, zgoubi computes field and derivatives along the element using a
field form factor . F = FE × FS . . FE (respectively . FS ) is the value of the Enge model
coefficient (Eq. 14.11) at distance .d E (resp. .d S ) from the entrance (resp. exit) EFB.
This may have the immediate effect, apparent in Fig. 14.4, that the integrated
field is not the expected value . B × L from the input data . L and . B, and may require
adjusting (increasing) . B so to recover the required . B dl.
14.4 Focusing 613

14.3.4 Toroidal Condenser

This is the ELCYLDEF element in zgoubi. With proper parameters, it can be used
as a spherical, a toroidal or a cylindrical deflector.
Motion along the optical axis, an arc of a circle of radius .r normal to electric field
.E, satisfies
p
. Er = v = v(Bρ)
q

with . p = mv the particle momentum, .q its charge and .(Bρ) = p/q the particle
rigidity.
The first order transport matrix of an electrostatic bend writes
⎛ ⎞
2−β 2
⎜
C x Sx 0 0 0 px2 0
r (1 − C x ) ⎟
⎜ ⎟
⎜ 2−β 2 ⎟
⎜ C x Sx 0 0 0 r 0 Sx ⎟
⎜ ⎟
⎜ 0 0 C y Sy 0 0 ⎟
⎜ ⎟
. Tcondenser = ⎜ ⎟ (14.16)
⎜
⎜
0 0 C y S y 0  0 ⎟
⎟
⎜  2 ⎟
⎜ 2−β 2 2 2−β 2 ⎟
⎜ − r Sx − 2−β r (1 − C x ) 0 0 1 r0 α
p2 0
1
γ2
− px2
(1 − Sx
r0 α ) ⎟
⎝ 0 x ⎠
0 0 0 0 0 1

⎡
α = deflection angle
⎢ C = cos pα
⎢
⎢ C  = dC = − p2 S
⎢
⎢ ds r2
.with ⎢ S =
r
sin pα
⎢  pd S
⎢ S = ds = C
⎢ 
⎣ (∗)x : p = px = 2 − β 2 − r0 /R0
√
(∗) y : p = p y = r0 /R0

14.4 Focusing

Particle beams are maintained confined along a reference propagation axis by means
of focusing techniques and devices. Methods available in zgoubi to simulate those
are addressed here.

14.4.1 Wedge Focusing

 A wedge angle .ε causes a particle at local

Wedge focusing is sketched in Fig. 14.5.
excursion .x to experience a change . B y ds = x B y tan ε in the field integral, com-
614 14 Optical Elements and Keywords, Complements

Fig. 14.5 Left: a focusing wedge (.ε < 0 by convention); opening the sector increases the horizontal
focusing. Right: a defocusing wedge (.ε > 0); closing the sector decreases the horizontal focusing.
The effect is the opposite in the vertical plane, opening/closing the sector decreases/increases the
vertical focusing

pared to the field integral through the sector magnet. In the linear approximation this
causes a change in trajectory angle

1 tan ε
. x = B y ds = x (14.17)
Bρ ρ0

with . Bρ the particle rigidity and .ρ0 its trajectory curvature radius in the field . B0
of the dipole. Vertical focusing results from the non-zero off-mid plane radial field
component
 . Bx in the fringe field region (Fig. 14.7): from (Maxwell’s equations)

∂
.
∂y
Bx ds = ∂∂x B y ds and Eq. 14.17 the change in trajectory angle comes out to
be 
 1 tan ε
. y = Bx ds = −y (14.18)
Bρ ρ0

A first order correction .ψ to the vertical kick accounts for the fringe field extent
(it is a second order effect for the horizontal kick):

tan(ε − ψ)
. y  = −y (14.19)
ρ0

with

λ 1 + sin2 ε B(s) (B0 − B(s))
.ψ = I1 with I1 = ds (14.20)
ρ0 cos ε edge λ B02

λ is the fringe field extent, . I1 quantifies the flutter (see Sect. 4.2.1); a longer/shorter
.
field fall-off (smaller/greater flutter) decreases/increases the vertical focusing.
14.4 Focusing 615

Fig. 14.6 Field components

in the . B y (s) field fall-off at a
dipole EFB

Fig. 14.7 Field components

off mid-plane, in the fringe
field region at the ends of a
dipole (. y > 0, here, referring
to Fig. 14.6). . B// parallel to
the particle velocity has no
effect. . Bx pulls a positively
charged particle away from
the median plane, under the
effect of a .v × Bx force
component. Inspection of the
. y < 0 region gives the same
result: the charge is pulled
away from the median plane

Linear approach
A wedge focusing first order transport matrix writes
⎛ ⎞
1 0 0 0 0 0
⎜ tanρ ε 1 0 0 0 0⎟
⎜ ⎟
⎜ 0 0 1 0 0 0⎟
T =⎜
⎜ 0 0 − tan ε
⎟ (14.21)
0⎟
. wedge
⎜ ρ
1 0 ⎟
⎝ 0 0 0 0 1 0⎠
0 0 0 0 0 1

Substitute .ε − ψ to .ε in the . R43 coefficient, when accounting for fringe field

extent .λ.
616 14 Optical Elements and Keywords, Complements

Fig. 14.8 Left: a quadrupole magnet [11]. Right: field lines and forces (assuming positive charges
moving out of the page) over the cross section of an horizontally focusing/vertically defocusing
quadrupole

14.4.2 Quadrupole

Quadrupoles are the optical lenses of charged particle beams, they ensure confine-
ment of the beam in the vicinity of the optical axis. Most of the time in beam lines and
cyclic accelerators, guiding and focusing are separate functions, focusing is assured
by quadrupoles, magnetic most frequently, possibly electrostatic at low energy.
The field in quadrupole lenses results from hyperbolic equipotentials, .V = ax y.
Pole profiles follow these equipotentials, in a .2π/4-symmetrical arrangement for
technological simplicity.

Magnetic Quadrupole

Magnetic quadrupoles are the optical lenses of high energy beams (Fig. 14.8).
The theoretical field in a quadrupole can be derived from Eq. 14.8 for the scalar
potential, with .n = 2 which yields

G  (X ) 2 G (4) (X ) 2
. V2 (X, Y, Z ) = G(X )Y Z − (Y + Z 2 )Y Z + (Y + Z 2 )2 Y Z − · · ·
12 384
(14.22)
and

∂ V2 G  (X ) 2
. B X (X, Y, Z ) = − = G  (X )Y Z − (Y + Z 2 )Y Z + · · · (14.23)
∂X 12
∂ V2 G  (X )
. BY (X, Y, Z ) = − = G(X )Z − (3Y 2 + Z 2 )Z + · · · (14.24)
∂Y 12
∂ V2 G  (X ) 2
. B Z (X, Y, Z ) = − = G(X )Y − (Y + 3Z 2 )Y + · · · (14.25)
∂Z 12
14.4 Focusing 617

Fig. 14.9 Horizontal and

vertical projections of
particle trajectories across a
stigmatic quadrupole
doublet. The first quadrupole
(QF) is horizontally focusing
(. K > 0; thus vertically
defocusing), the second one
(QD) has reverse sign
(. K < 0) and reverse effect

. G(X ) is given by Eq. 14.4 whereas

B0 G0
. G0 = and K = (14.26)
a Bρ

define respectively the quadrupole gradient and strength, the latter relative to the
rigidity . Bρ. The quadrupole is horizontally focusing and vertically defocusing if
. K > 0, and the reverse if . K < 0, this is illustrated in Fig. 14.9 which shows the
effect of a doublet of quadrupoles with focusing strengths of opposite signs.
Linear approach
The first order transport matrix of a quadrupole with length. L, gradient.G and strength
K = G/Bρ writes
.

⎛ ⎞
Cx Sx 0 0 00 ⎡ √
⎜ C x
⎜ Sx 0 0 00 ⎟
⎟ C x = cos L K ; C x = dC d L = −K Sx
x
⎢ √
⎜ 0 0 Cy Sy 00 ⎟ ⎢ S = √1 sin L K ; S  = d Sx = C
⎜ ⎟ x x
⎟ with ⎢
x dL
. Tquad = ⎜ C y S y
K √
⎜ 0 0 00 ⎟ ⎢  dC y (14.27)
⎜ ⎟ ⎢ C y = cosh L K ; C y = d L = K S y
⎜ 0 L ⎟ ⎣ √
0 0 0 1 2 dS
⎝ γ ⎠ S y = √1 sinh L K ; S y = d Ly = C y
K
0 0 0 0 0 1

. K > 0 for a focusing quadrupole (by convention, in the.(x, x  ) plane, thus defocusing
in the .(y, y  ) plane). Permute the horizontal and vertical .2 × 2 sub-matrices in the
case of a defocusing quadrupole.

Electrostatic Quadrupole

The hypotheses are those of Sect. 2.2.2: paraxial motion, field normal to velocity, etc.
Take the notations of Eqs. 2.25 and 2.26 for the field and potential, case of electrodes
618 14 Optical Elements and Keywords, Complements

Fig. 14.10 A sketch of a

solenoid, and quantities used
to define it

in the horizontal and vertical planes (Fig. 2.14). Electrode potential is .±V /2, pole
tip radius .a, so that . K = −V /2a 2 in Eq. 2.26. The equations of motion then write
⎡
d2 x
+ Kx x = 0 −q V V 1
.⎣
ds 2
with K x = −K y = =± 2 (14.28)
d y 2
+ Ky y = 0 a 2 mv2 a |Eρ|
ds 2   
electrical
rigidity

The transport matrix is the same as for the magnetic quadrupole, Eq. 14.27, taken for
that . K value.

14.4.3 Solenoid

Assume a solenoid magnet with longitudinal axis (OX). In a cylindrical frame

(O; X, r, φ), Fig. 14.10 (.r is the radial coordinate, the angle.φ is taken in the X-normal
.
plane), . Bφ (X, r, φ) ≡ 0. Take solenoid length . L, mean coil radius .r0 and an asymp-
totic field . B0 = μ0 N I /L, with . N I = number of ampere-Turns, .μ0 = 4π × 10−7
H/m. The asymptotic field value is defined by
 ∞
. B X (X, r < r0 ) d X = μ0 N I = B0 L independent of r (14.29)
−∞

There is a variety of methods to compute the field vector .B(X, r ). Opting for one
in particular may be a matter of compromise between computing speed and field
modeling accuracy. A simple model is the on-axis field
⎡ ⎤
B0 ⎣ L/2 − X L/2 + X ⎦
. B X (X, r = 0) =  + (14.30)
2 (L/2 − X )2 + r02 (L/2 + X )2 + r02
14.4 Focusing 619

with . X = r = 0 taken at the center of the solenoid. This model assumes that the coil
thickness is small compared to its mean radius .r0 . The magnetic length comes out to
be
∞ 
−∞ B X (X, r < r 0 )d X 4r 2
. L mag ≡ = L 1 + 20 > L (14.31)
B X (X = r = 0) L

so satisfying

μ0 N I r0 XL μ0 N I
.on-axis B X (X = r = 0) = −−−−→
4r02 L
L 1+
L2

Maxwell’s equations and Taylor expansions provide the off-axis field .B(X, r ) =
(B X (X, r ), Br (X, r )). One has in particular in the .r0 X L limit,

μ0 N I −r d B X
. B X (X, r ) = and Br (X, r ) = (14.32)
L 2 dX

An other way to compute the field vector .B(X, r ) is the elliptic integrals technique
developed in [12], which constructs . B X (X, r ) and . Br (X, r ) from respectively
 
μ0 N I ck r0 − r
. B X (X, r ) = X K+ ( − K ) (14.33)
4π r 2r0
1 r0 ! "
Br (X, r ) = μ0 N I 2(K − E) − k 2 K
k r

wherein . K , . E and . are the three complete elliptic integrals, .X is an . X - and . L-

dependent form factor, and
√  √
.k = 2 r0 r / (r0 + r )2 + X 2 ; c = 2 r0 r /(r0 + r )

As an illustration, Fig. 14.11 displays a trajectory across a . L = 1 m solenoid and

its field fall-offs, and the field experienced along that trajectory, in the axial model of
Eq. 14.30. In the paraxial approximation, a pitch requires a distance .l = 2π/K , with
. K = B0 /Bρ the solenoid strength, which is a condition satisfied here if the fringe
field extent is short enough (solenoid radius .r0 is small enough).
Linear approach
The equations of motion write, to the first order in the coordinates, in respectively
the central region (field . Bs ) and at the ends (at .s = sEFB ),

   K
 x  − K z  = 0 x − z δ(s − sEFB ) = 0
  2
. and  (14.34)
 z  + K x  = 0   K
z + x δ(s − sEFB ) = 0
2
620 14 Optical Elements and Keywords, Complements

Fig. 14.11 Left: Horizontal (Y) and vertical (Z) projections of a particle trajectory across a. L = 1 m
solenoid, with additional 1 m extents upstream and downstream of the coil to account for the
extended field fall-offs. The particle is launched with zero incidence, from transverse position
.Y = Z = 0.5 mm. Sample solenoid radius/length values in the range .0.001 ≤ r 0 /L ≤ 0.2 show
that only for smallest .r0 /L = 0.001 does the trajectory end with .Y = Z = 0.5 mm and quasi-zero
incidence (the thicker Y(X) and Z(X) curves), whereas greater .r0 /L causes final Y(X) and Z(X) to
be offset. Right: field . B X (X, r ) experienced along the trajectory for the various .r0 /L values, the
steep fall-off case is for .r0 /L = 0.001

The first order transport matrix of a solenoid with length . L writes

⎛ 2 2 2
⎞
C2 K SC SC KS 0 0
⎜ −K
SC C 2 − K2 S 2 SC 0 0 ⎟
⎜ 2 ⎟ ⎡
⎜ −SC − K2 S 2 C 2 2 ⎟ K = BρBs
⎜ K SC 0 0 ⎟
. Tsol =⎜
⎜
K 2
−SC − 2 SC C 2 0 0
K ⎟ with ⎣ C = cos K L
⎟ (14.35)
2 S 2
⎜ L ⎟ S = sin K2L
⎜ 0 0 0 0 1 2 ⎟
⎝ γ ⎠
0 0 0 0 0 1

A solenoid rotates the decoupled axis longitudinally by an angle .α = K L/2 =

Bs L/2Bρ.

14.5 Data Treatment Keywords

14.5.1 Concentration Ellipse: FAISCEAU, FIT[2],

MCOBJET, …

It is often useful to associate the projection of a particle bunch in the horizontal,

vertical or longitudinal phase space with an rms phase space concentration ellipse
(CE). Various keywords in zgoubi resort to concentration ellipses:

– FAISCEAU for instance prints out, in zgoubi.res, CE parameters drawn from

particle coordinates,
– random particle distributions by MCOBJET are defined using CE parameters,
14.5 Data Treatment Keywords 621

– ellipse parameters computed from CEs are possible constraints in FIT[2] proce-
dures.
Transverse phase space graphs by zpop also compute CEs.
The CE method is resorted to in various exercises, for instance for comparison
of the ellipse parameters it gets from the rms matching of a bunch, with theoretical
beam parameters derived from first order transport formalism (such as computed
from rays by MATRIX, or TWISS).
The CE method used in these various keywords and data treatment procedures is
the following. Let .z i (s), z i (s) be the phase space coordinates of .i = 1, n particles
in a set observed at some azimuth .s along an optical sequence. The second moments
of the particle distribution are

1
n
z (s) =
. 2 (z i (s) − z(s))2
n i=1
1
n
zz  (s) = (z i (s) − z(s))(z i (s) − z  (s)) (14.36)
n i=1
1 
n
z 2 (s) = (z (s) − z  (s))2
n i=1 i

From these, a concentration ellipse is defined, encompassing a surface . Sz (s), with

equation
 2
.γc (s)z + 2αc (s)zz + βc (s)z = Sz (s)/π
2
(14.37)
2
Noting . = z 2 (s) z 2 (s) − zz  (s), the ellipse parameters write

z 2 (s) zz  (s) z 2 (s) √

.γc (s) = √ , αc (s) = − √ , βc (s) = √ , Sz (s) = 4π (14.38)

With these conventions, the rms values of the .z and .z  projected densities satisfy

Sz Sz
σ =
. z βz and σz  = γz (14.39)
π π

14.5.2 Transport Coefficients: MATRIX, OPTICS, TWISS,

Etc.

Zgoubi does not know about matrix transport, it does not define optical elements by a
transport matrix, it defines them by electrostatic and/or magnetic fields in space (and
time possibly). Well, except for a couple of optical elements, for instance TRANS-
MAT, which pushes particle coordinates using a matrix, or SEPARA, an analytical
622 14 Optical Elements and Keywords, Complements

mapping through a Wien filter. Zgoubi does not transport particles using matrix prod-
ucts either, it does that by numerical integration of Lorentz force equation through
these .E and/or .B fields.
However it is often useful to dispose of a matrix representation of an optical
element or a beam line, or of paraxial parameters drawn from the first or second order
one-turn mapping of a ring accelerator. Several commands in zgoubi perform the
required treatment to derive these informations from particle coordinates. Examples
are MATRIX: computation of matrix transport coefficients up to 3rd order, from
initial and current coordinates of a particle sample. OPTICS transports a beam matrix,
given its initial value using OBJET[KOBJ .= 5.1]. TWISS derives a periodic beam
matrix from a 1-turn mapping of a periodic sequence, and transports it from end to
end so generating the optical functions along the sequence.
These capabilities are resorted to in the exercises. It may be required for instance
to compare transport coefficients derived from raytracing, with the matrix model
of the optical element(s) concerned. Or to compute a periodic beam matrix in a
periodic optical sequence, this is how betatron functions are produced, often for the
mere purpose of comparisons with matrix code outcomes, or with expectations from
analytical models.

Coordinate Transport

In the Gauss approximation (i.e., trajectory angle .θ ∼ sin θ ), particles follow paths
which can be described with simple functions: parabolic, sinusoidal or hyperbolic.
A consequence is that a string of optical elements, and coordinate transport through
the latter, can be handled with a simple mathematics toolbox. Taylor expansion (also
known as transport) techniques are part of it, whereby a coordinate excursion .v2i
(with index .i = 1 → 6 standing for .x, x  , y, y  , δs or .δp/ p) from some reference
trajectory at a location .s2 along the line is obtained from the excursions . v1i at an
upstream location .s1 , via


6 
6 
6
v =
. 2i Ri j v1 j + Ti jk v1 j v1k + v1i jkl v1 j v1k v1l + ... (14.40)
j=1 j,k=1 j,k,l=1

This Taylor development can be written under matrix form, for instance to the first
order in the coordinates, for non-coupled motion,
⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
x T11 T12 0 0 0 T16 x x
⎜ x ⎟ ⎜ T21 T22 0 0 0 ⎟ ⎜
T26 ⎟ ⎜ x ⎟ ⎟ ⎜  ⎟
⎜ ⎟ ⎜ ⎜ x ⎟
⎜ y ⎟ ⎜ 0 0 T33 T34 0 T36 ⎟ ⎜ y ⎟ ⎜ y ⎟
.⎜ ⎟ ⎜
⎜ y ⎟ = ⎜ 0
⎟⎜ ⎟ ⎜
⎜  ⎟ = T (s2 ← s1 ) ⎜ y  ⎟
⎟ (14.41)
⎜ ⎟ ⎜ 0 T43 T44 0 T46 ⎟
⎟⎜ y ⎟ ⎜ ⎟
⎝ δs ⎠ ⎝ 0 0 0 0 T55 T56 ⎠ ⎝ δs ⎠ ⎝ δs ⎠
δp/ p 2 0 0 0 0 T65 T66 δp/ p 1 δp/ p 1
14.5 Data Treatment Keywords 623

These are the quantities which such keywords as MATRIX [1, cf. Sect. 6.5] and
OPTICS [1, cf. Sect. 6.4] compute, from particle coordinates. Most of the time they
are resorted to for mere comparison with theoretical matrices such as recalled in
Sects. 14.2–14.4.

Beam Matrix

OPTICS and TWISS keywords cause the transport of a beam matrix. The former
requires initial beam ellipse parameters: these are provided as part of the initial
object definition, by OBJET. The latter first derives a periodic beam matrix from
initial and final particle coordinates resulting from raytracing throughout an optical
sequence. Basic principles are recalled here, regarding the way these keywords work
in zgoubi. They are resorted to quite often in the exercises.
In the linear approximation, the transverse phase space ellipse associated with a
particle distribution (for instance, the concentration ellipse, Sect. 14.5.1) is written
(with .z standing for indifferently .x or . y)
εz
γ (s)z 2 + 2αz (s)zz  + βz (s)z 2 =
. z (14.42)
π
in which the ellipse parameters

1 dβz 1 + α2
β (s), αz (s) = −
. z , γz (s) = (14.43)
2 ds βz

are functions of the observation location .s along the optical sequence. The surface
ε of the ellipse is an invariant if the beam travels in magnetic fields, however field
. z

non-linearities, phase space dilution, etc. may distort the distribution and change the
surface of its rms matching concentration ellipse. In the presence of acceleration or
deceleration the invariant quantity is .βγ εz instead, with .β = v/c and .γ the Lorentz
relativistic factor.
The ellipse Eq. 14.42 can be written under the matrix form
 
z
[z, z  ] σz−1 (s) =1 (14.44)
z
.

with .σz the beam matrix:

εz βz (s) −αz (s)

σ =
. z (14.45)
π −αz (s) γz (s)

The ellipse parameters can be transported from .s1 to .s2 using

σ
. z,2 = T σz,1 T̃ (14.46)
624 14 Optical Elements and Keywords, Complements

with .T = T (s2 ← s1 ) the transport matrix (Eq. 14.41) and .T̃ its transposed. This can
also be written under the form
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
βz 2
T11 −2T11 T12 2
T12 βz
. ⎝ αz ⎠ = ⎝ −T11 T21 T21 T12 + T11 T22 −T12 T22 ⎠ ⎝ αz ⎠ (14.47)
γz 2 2
T21 −2T21 T22 2
T22 s2 ←s1
γz 1

(subscripts 1, 2 normally hold for horizontal plane motion, .z = x: change to 3, 4

for vertical motion, .z = y). This beam matrix formalism can be extended to the
longitudinal phase space and coordinates (.δs, .δp/ p). Thus a .6 × 6 beam matrix can
be defined, ⎛ ⎞
σ11 σ12 0 0 0 σ16
⎜ σ21 σ22 0 0 0 σ26 ⎟
⎜ ⎟
⎜ 0 0 σ33 σ34 0 σ36 ⎟
.σ = ⎜ ⎟ (14.48)
⎜ 0 0 σ43 σ44 0 σ46 ⎟
⎜ ⎟
⎝ 0 0 0 0 σ55 σ56 ⎠
0 0 0 0 σ65 σ66

This can be generalized to non-zero anti-diagonal terms, if motions are coupled.

Periodic Structures

In the hypothesis of an . S- periodic structure: a long beam line with repeating pattern,
a cyclic accelerator, transverse motion stability requires the transport matrix over a
period, from .s to .s + S to satisfy

[Ti j ](s + S ← s) = I cos μ + J sin μ

. (14.49)

where .μ = (S) ds/β is the betatron phase advance over the period (independent of
the origin),

10 αz (s) βz (s)
. I = is the identity matrix, J = (and J 2 = −I )
01 −γz (s) −αz (s)
(14.50)

14.6 Exercises

14.1 Magnetic Sector Dipole

Solution 14.1.
(a) Simulate a .ρ = 0.5 m radius, .α = 60◦ sector dipole with n .= −0.6 field index,
in both cases of hard edge and of soft fall-off fringe field model. Find the reference
14.6 Exercises 625

Fig. 14.12 Symmetric point to point focusing, case of a 60.◦ or a 180.◦ sector dipole


arc, such that . arc B ds = B L with . L the arc length in the hard-edge model and B
the field along that arc.
Make sure that the reference arc has the expected length.
Produce the field along the reference arc, for a few different values of the fringe-
field extent.
(b) A possible check of the first order: OBJET[KOBJ .= 5], MATRIX[IORD .=
1, IFOC .= 0] can be used to compute the transport matrix from the rays. Compare
what it gives with theory.
(c) Consider a sector dipole with parallel gap, uniform field. Show the well known
geometrical property of point-to-point focusing represented in Fig. 14.12.
Produce the aberration curve .x  (x) in the horizontal phase-space at the image
plane.
Test the convergence of the numerical solution versus integration step size.
(d) Transport a proton along the reference axis, injected with its spin tangent to
the axis. Compare spin rotation with theory.
Test the convergence of the numerical solution versus integration step size.

14.2 Quadrupole Doublet

Solution 14.2.
Reproduce Fig. 14.9.

14.3 Solenoid
Solution 14.3.
An introduction to SOLENOID.
(a) Reproduce Fig. 14.11. Use both field models of Eqs. 14.30 and 14.33 and com-
pare their outcomes, including the first order paraxial transport matrices, and some
higher order coefficients as well (computed from in and out trajectory coordinates).
(b) Compare final coordinates in (a) with outcomes from the first order transport
formalism (Sect. 14.4.3).
(c) Make a 1-dimensional (on-axis) field map of a .r0 = 10 cm, . L = 1 m solenoid
(namely, a map . BX,i (X i ) of the field at the nodes of a X-mesh with mesh size . X i+1 −
X i ). Reproduce the trajectory in (a) (case .r0 = 10 cm) using that field map, with the
keyword BREVOL. Check the convergence of the final particle coordinates, using
the field map, depending on the mesh size.
626 14 Optical Elements and Keywords, Complements

14.7 Solutions of Exercises of This Chapter: Optical

Elements and Keywords, Complements

14.1 Magnetic Sector Dipole

DIPOLE input data.
(a) A simulation of a.ρ = 0.5 m radius, 60.◦ sector dipole with n.= −0.6 field index,
in the hard-edge field model, is given in Table 14.1. A simulation which includes
fringe fields is given in Table 14.2.
A major difference between the two is in the angular extent of the field domain,
AT, in order to allow encompassing the fringe field extents, however there is more,
as follows.

Hard edge model

The effective field boundaries (EFB) have to be placed on the angular opening lim-
its, which means, in the representation of Fig. 14.13, and according to the users’
guide [13, see DIPOLE],

ω+ = AC E N T > 0, ω− = −AT + AC E N T < 0, ω+ − ω− = AT > 0

.

Otherwise, in the case AT would be greater than the magnet deflection angle.α = 60◦ ,
particles would jump from zero field to plateau field value over the EFB, and so miss
part of the field integral. Note that for mere code-specific, geometry computation
reasons, it also requires that ACENT = AT/2, so that, in fine, .ω+ = −ω− = AT /2.

Table 14.1 Input data file: definition of a dipole with index in the hard-edge field model. Definition
of the [#S_60dSectDip_hardE:#E_60dSectDip_hardE] segment, mostly for the purpose of possible
further INCLUDE. This file is used under the name sectorDIP_hardE.inc in subsequent exercises
14.7 Solutions of Exercises of This Chapter: Optical Elements … 627

Table 14.2 Input data file: definition of a dipole with index in the soft-edge field model. The field
extent in the Enge model (Eq. 14.11) is taken to be .g = 5 cm (.λ E = λ S = g in Users’ Guide’s
notations), so subtended by an angle .atan(g/R M) = 5.71059◦ , thus well comprised in a 10.◦
angular aperture. ACENT value is free, 30.◦ as adopted here is arbitrary, it is just left to the
value it was given in the hard edge settings (Table 14.1). This input includes the definition of
the [#S_60dSectDip_softE:#E_60dSectDip_softE] segment. This file is used under the name sec-
torDIP_softE.inc in subsequent exercises

Fig. 14.13 Parameters used

to define the geometry of a
dipole magnet with index,
using DIPOLE [13,
see DIPOLE]

Soft edge model

AT has to be greater than the magnet deflection angle .α = 60◦ in order to encompass
the fringe field extent beyond the entrance and exit EFBs, so that, in the representation
of Fig. 14.13, and according to the users’ guide,
628 14 Optical Elements and Keywords, Complements

. AC E N T > ω+ , |ω− | < AT − AC E N T

Integration-wise, particles will smoothly traverse the field fall-off regions, step by
step, no field discontinuity there. Note that motion integration accuracy requires the
step size to be small enough, compared to the fringe field extent. In the notations of
Fig. 14.13, the resulting additional optical axis lengths .l E and .l S within the AT sector,
on entrance and exit side respectively, to account for the field fall-offs, write

l = R M × tan(AC E N T − ω+ ),
. E l S = R M × tan[AT − (AC E N T − ω− )]

Checking back one fortunately finds

lE + lS
. atan + ω
 − ω− + atan = AT
RM RM
   magnet body   
entrance exit
fringe field fringe field

It also results from the fringe field modeling that the reference trajectory (which
is ideally the trajectory that coincides with R .= RM in the body of the magnet)
enters the AT sector at radius RE, with an incidence TE. These two quantities have
to be accounted for in setting the entrance and exit reference frames, however this is
user’s matter, regarding the choice of reference frames: most often (in synchrotron
rings for instance) the reference curve is R .= RM, so that Y and T coordinates of
the reference particle are zero (the moving frame has its origin at the origin of the
polar frame in which the field is defined, and rotates with the particle, clockwise in
Fig. 14.13 representation). Thus, one has to set

. T E = −(AC E N T − ω+ ) < 0, R E = R M/ cos T E

Note that, because of the small deflection due to fringe fields, RS and TS need be
adjusted if the DIPOLE process has to end up with the reference particle featuring
zero Y and T coordinates. Expectedly, that would be satisfied with RS and TS values
near
−
. T S = AT − (AC E N T − ω ) > 0, RS = R M/ cos T S

The radius . R of the reference arc, such that . arc B ds = B L with . L the arc length
in the hard-edge model, has to be found. Same thing for the arcs at.±0.1% momentum
offset. FIT can be used for that (Table 14.3).

(b) First order transport.

This is left to the reader. Theoretical matrices are given in Eqs. 14.6 and 14.7.
Refer to exercises in earlier chapters, such comparison is often performed.
14.7 Solutions of Exercises of This Chapter: Optical Elements … 629

Table 14.3 Input data file: find closed orbits, using FIT or FIT2, and log stepwise data in
zgoubi.plt. Closed orbits are found for the reference particle (a particle with rigidity . Bρ =
5[kG] × 50[cm] kG cm) and for particles with .±δp/ p momentum offset. FIT starts with initial
.Y0 radius values resulting from a hard edge model, i.e., .Y0 = Bρ/B = 250[kG cm] /5[kG] and
.±0.1%. This file produces the field along these trajectories, an effect of DIPOLE[IL .= 2]. The
[#S_60dSectDip_softE:#E_60dSectDip_softE] segment of Table 14.2 is INCLUDEd; simply sub-
stitute [#S_60dSectDip_hardE:#E_60dSectDip_hardE] (as defined in Table 14.1) to work with the
hard edge model instead

(c) Point-to-point focusing.

The hard-edge model DIPOLE of Table 14.1 can be used, with the following
modifications and addenda in order to simulate the symmetric 60.d́eg sector and
drifts configuration of Fig. 14.12:
– add OBJET[KOBJ .= 1, IMAX .= 41] so to generate 41 particles launched with
T ∈ [−20, 20] mrad, like so:
. 0

’OBJET’
64.62444403717985
1
1 41 1 1 1 1
0. 1. 0. 0. 0. 0.
50. 0. 0. 0. 0. 3.8685052339

– following OBJET add a drift with length . R M/ tan(30◦ ) = 86.6025403784 cm,

– following DIPOLE add a drift with length . R M/ tan(30◦ ),
– in DIPOLE: set the field index to zero,
– add AUTOREF[I .= 3, I1 .= 1, I2 .= 2, I3 .= 3] after DIPOLE: that will cause
computation of the location of the waist formed by particles 1, 2 and 3,
– add FAISTORE[FNAME .= zgoubi.fai, IP .= 1] after AUTOREF, before END.
This logs particle data at that location.
630 14 Optical Elements and Keywords, Complements

Fig. 14.14 Aberration curve at the focal point of a .180◦ uniform field dipole: a second order
(sextupole) aberration, .Y ∝ T 2 , typical of a bend non-linearities

In the execution listing zgoubi.res one finds:

This indicates that AUTOREF confirms expectations: it found the waist

– at . XC = 0, which means right at the end of the downstream drift,

– at a radial excursion.Y C = 50 cm as expected (the origin of the Y axis is at DIPOLE
curvature center),
– with the reference frame X axis at an angle . A = 0 to particle 1 direction of motion.

QED.
The following gnuplot script can be used to print the horizontal phase space .T (Y )
at the image plane (Fig. 14.14)

In the case of an.α = 180◦ dipole, the previous input data file can be used, changing
DIPOLE angles to . AT = ω+ − ω− = 180◦ with for instance .ω+ = −ω− = 90◦ .
Remove the drifts in order to obtain the .180◦ sector configuration of Fig. 14.12.
Step size:
The method is the same as in Exercise 2.2 (b), case of a toroidal condenser, which
can be referred to.
14.7 Solutions of Exercises of This Chapter: Optical Elements … 631

(d) Spin precession.

Add SPNTRK[KSO .= 1] at the beginning of the input data file to track spin,
starting aligned on the . X axis. Tracking spin also requires PARTICUL, in order to
define particle’s mass, charge and anomalous magnetic moment.
The theoretical value of the spin precession angle in the moving frame is .Gγ α
(Eq. 3.32), with .α = π/3 or .α = π in the previous two deflection cases considered.
This is the value which the stepwise integration produces.

14.2 Quadrupole Doublet

The input data file for this problem is given in Table 14.4.

Table 14.4 Input data file: a double-focus quadrupole doublet

632 14 Optical Elements and Keywords, Complements

Table 14.5 Input data file: a 1 m long solenoid, with 1 m upstream and downstream fringe field
extents. The initial coil radius is .r0 = 0.1 cm, it is scanned (by REBELOTE) over the range .1 ≤
r0 ≤ 20 cm. For each.r0 a particle is launched with initial position.Y = Z = 1 mm and initial angles
.T = P = 0

14.3 Solenoid

(a) The paraxial trajectory pitch is .l = 2π Bρ/B0 (Sect. 14.4.3). Take . L = 1 m

(Fig. 14.11) and . Bρ = 1 T m for simplicity, thus . B0 = 2π T. Assume a particle
launched from .Y = Z = 1 mm with zero incidence. Scan the solenoid radius value
in the range .1 ≤ r0 ≤ 200 mm to reproduce the figure. The data to be plotted
(. X, Y, Z , B X ) are read from zgoubi.plt.
The beam optics model is given in Table 14.5. Note the use of KOBJ .= 2 in
OBJET, which allows creating particles in an arbitrary number (just one, here), with
arbitrary initial coordinates. REBELOTE[IOPT .= 1] is used to repeat the sequence,
varying the parameter . R0 under SOLENOID.
(b) To allow comparison, theoretical matrices (Eq. 14.35) must be computed for the
theoretical length, L, of the matrix transport solenoid model. Tracking must extend
upstream and downstream of the solenoid, over a distance much greater than the
solenoid diameter (the latter determines the field fall extent, Eq. 14.30) (Table 14.7).
(c) A 1-dimensional (on-axis) field map of the solenoid field,. B X,i (X i ), can simply
be generated by tracking a particle along the solenoid axis. It has to extend upstream
and downstream of the solenoid, over a distance much greater than the solenoid
diameter. The integration step size will be the mesh size, take it in the centimeter
range (. r0 ), 5 cm here. An intermediate stage is necessary, which consists in reading
. X, B X (X ) from zgoubi.plt and re-writing it in a dedicated ASCII file in a format

proper for use by the keyword BREVOL.

The input file to generate the field and log to zgoubi.plt is given in Table 14.6.
Similar exercises, generating a 1D field map and using BREVOL, can be found
be found in zgoubi sourceforge repository [14].
References 633

Table 14.6 Input data file: track a particle along the central axis of the solenoid, to generate a 3 m
long, 1D field map, with mesh step 5 cm

Table 14.7 Input data file: track a particle in the solenoid, in a similar manner to the input data
file of Table 14.6, using a field map model instead

References

1. Zgoubi Users’ Guide, updated Sourceforge version (at revision 2037, here): https://
sourceforge.net/p/zgoubi/code/HEAD/tree/trunk/guide/Zgoubi.pdf. Méot, F.: Zgoubi Users’
Guide. Report BNL-98726-2012-IR, C-A/AP 470 (2012). https://www.osti.gov/servlets/purl/
1062013
2. The AGS at the Brookhaven National Laboratory. https://www.bnl.gov/rhic/AGS.asp
3. The CERN PS. https://home.cern/science/accelerators/proton-synchrotron
4. J.T. Volk, Experiences with permanent magnets at the Fermilab recycler ring. James T Volk 2011
JINST6 T08003. https://iopscience.iop.org/article/10.1088/1748-0221/6/08/T08003/pdf
5. F. Méot, F. Lemuet, Developments in the ray-tracing code Zgoubi for 6-D multiturn tracking
in FFAG rings. NIM A 547, 638–651 (2005)
6. H.A. Enge, Deflecting magnets, in Focusing of Charged Particles, vol. II ed. by A. Septier
(Academic Press Inc., 1967), pp. 203–264
7. Y. Dutheil, et al., A model of the AGS based on stepwise ray-tracing through the measured
field maps of the main magnets, in Proceedings of IPAC2012, New Orleans, Louisiana, USA,
TUPPC101, pp. 1395–1399. https://accelconf.web.cern.ch/IPAC2012/papers/tuppc101.pdf; F.
Méot, et al., Modeling of the AGS using zgoubi - status, in Proceedings of IPAC2012, New
Orleans, Louisiana, USA, MOPPC024, pp. 181–183. https://accelconf.web.cern.ch/IPAC2012/
papers/moppc024.pdf
634 14 Optical Elements and Keywords, Complements

8. R.E. Thern, E. Bleser, The dipole fields of the AGS main magnets, BNL-104840-2014-TECH,
1/26/1996. https://technotes.bnl.gov/PDF?publicationId=31175
9. F. Méot, L. Ahrens, K. Brown, et al., A model of polarized-beam AGS in the ray-tracing
code Zgoubi. BNL-112453-2016-TECH, C-A/AP/566 (2016). https://technotes.bnl.gov/PDF?
publicationId=40470, https://www.osti.gov/biblio/1336073
10. G. Leleux, Accélérateurs Circulaires. Lectures at the Institut National des Sciences et Tech-
niques du Nucléaire, CEA Saclay (1978). (unpublished)
11. Credit: Brookhaven National Laboratory. https://www.flickr.com/photos/brookhavenlab/
8495311598/in/album-72157611796003039/
12. M.W. Garrett, Calculation of fields [...] by elliptic integrals. J. Appl. Phys. 34(9) (1963)
13. F. Méot, Zgoubi Users’ Guide. https://www.osti.gov/biblio/1062013-zgoubi-users-guide
Sourceforge revision 1379 (2020-02-29). https://sourceforge.net/p/zgoubi/code/HEAD/tree/
trunk/guide/Zgoubi.pdf
14. https://sourceforge.net/p/zgoubi/code/HEAD/tree/branches/exemples/KEYWORDS/
BREVOL/

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0
International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing,
adaptation, distribution and reproduction in any medium or format, as long as you give appropriate
credit to the original author(s) and the source, provide a link to the Creative Commons license and
indicate if changes were made.
The images or other third party material in this chapter are included in the chapter’s Creative
Commons license, unless indicated otherwise in a credit line to the material. If material is not
included in the chapter’s Creative Commons license and your intended use is not permitted by
statutory regulation or exceeds the permitted use, you will need to obtain permission directly from
the copyright holder.
Index-Zgoubi Optical Elements

A ELMULT, 597
AUTOREF, 42, 221, 461, 472, 588, 629 ERRORS, 369, 372

B F
BEND, 358, 510, 517, 594 FAISCEAU, 81, 99–101
BREVOL, 633 found in most exercises, e.g., 130, 424, 585

FAISCNL, 524, 528

C FAISTORE, 93, 98, 114, 125
CAVITE, 75, 78, 93, 94, 96, 122, 123, 125, found in most exercises, e.g., 90, 423, 601
128, 167, 171, 197, 217, 234, 362, 372,
378, 423, 466, 472 FFAG, 409
CHAMBR, 549, 593 FFAG-SPI, 428
CHANGREF, 462, 585 FINISH, 221, 461, 472
COLLIMA, 46, 549 FIT[2], 84, 90, 98, 109, 115, 119, 129, 154,
CYCLOTRON, 178 165, 169, 170, 174, 178, 232, 358, 364,
369, 412, 415, 424, 431, 433, 521, 527,
530, 552, 557, 560, 629
D
DIPOLE, 88, 90, 91, 111, 114, 116, 118, 121,
174, 197, 217, 585, 594, 626 G
cyclotron, 88 GOTO, 221, 461, 472
DIPOLE-M, 585
DRIFT, 40, 46, 129, 369, 462, 467, 510, 517,
521, 527, 530, 532, 534, 543, 594 H
split, 374, 510, 521, 543 HISTO, 129, 518, 524, 532, 534

E I
EBMULT, 521, 530, 535 IMAGE[S, Z], 43, 524, 528, 535, 585
EL2TUB, 49 INCLUDE, 46, 82, 84, 86, 93, 94, 98, 104,
ELCYLDEF, 43 109, 110, 114, 115, 117, 118, 122, 123,
ELMIR, 46 128, 158, 167, 170, 197, 203, 220, 234,

© The Editor(s) (if applicable) and The Author(s) 2024 635

F. Méot, Understanding the Physics of Particle Accelerators,
Particle Acceleration and Detection,
https://doi.org/10.1007/978-3-031-59979-8
636 Index-Zgoubi Optical Elements

359, 364, 369, 372, 378, 412, 415, 424, Q

431, 433, 437, 438, 466, 468, 472, 513, QUADRUPOL, 359, 510, 517, 585
517, 524, 527, 535, 543, 549, 557, 562,
590, 599, 629
R
REBELOTE, 41, 43, 47, 84, 90, 93, 94, 97,
M 98, 104, 110, 115, 117, 118, 120, 122,
MATRIX, 116, 118, 158, 169, 170, 358, 415, 123, 125, 128, 154, 165, 167, 170, 174,
424, 433, 462, 467, 510, 521, 527, 530, 178, 197, 203, 217, 232, 362, 368, 372,
543, 590, 594 378, 412, 415, 424, 431, 433, 437, 438,
MCDESINT, 532, 534, 549 462, 468, 517, 532, 534, 535, 546, 554,
MCOBJET, 125, 220, 362, 372, 425, 466, 560, 627, 632
470, 472, 517, 524, 528, 532, 534, 535,
549, 562, 601
MULTIPOL, 359, 368, 374, 378, 462, 467, S
510, 517, 543 SCALING, 197, 203, 234, 362, 364, 372,
378, 423, 468, 510, 513, 517
SOLENOID, 632
O SPINR, 104, 105
OBJET, 81 SPNPRT, 41, 97, 98, 100, 105, 552, 560, 562
(or MCOBJET) found in any input data
file, e.g., 40, 43, 629 SPNTRK, 41, 97, 102, 104, 105, 169, 171,
OPTICS, 513, 546 372, 378, 470, 552, 554, 560, 562
OPTIONS, 468, 513, 517, 530, 535, 543, 590 SRLOSS, 203, 372, 470
SRPRNT, 203, 372, 470
.plt, 47, 220, 234, 358, 364, 365, 378, 437 SVDOC, 369
SYSTEM, 41, 43, 46, 81, 85, 90, 94, 98, 104,
CONSTY, 73, 409, 428, 581 109, 115, 117, 122, 123, 126, 128, 154,
WRITE, 197, 203, 369 165, 167, 170, 174, 178, 217, 232, 360,
ORDRE, 43 368, 369, 372, 378, 409, 412, 415, 423,
428, 431, 433, 468, 513, 517, 532, 534,
535, 546, 549, 552, 554, 560, 590
P
PARTICUL, 40, 43, 46, 90, 94, 97, 98, 101,
104, 115, 122, 123, 128, 165, 167, 170, T
178, 197, 203, 217, 232, 362, 372, 378, TOSCA, 79, 87, 92, 109, 154, 165, 166, 232
412, 424, 428, 431, 462, 470, 472, 521,
527, 530, 532, 534, 535, 549, 552, 554, TWISS, 46, 199, 360, 364, 418, 435
560, 562, 593, 597
PICKUPS, 423
W
WIENFILTER, 51, 552, 554, 557, 560